fs/ext4/super.c

   1 /*
   2  *  linux/fs/ext4/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/fs/minix/inode.c
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  *
  15  *  Big-endian to little-endian byte-swapping/bitmaps by
  16  *        David S. Miller (davem@caip.rutgers.edu), 1995
  17  */
  18
  19 #include <linux/module.h>
  20 #include <linux/string.h>
  21 #include <linux/fs.h>
  22 #include <linux/time.h>
  23 #include <linux/vmalloc.h>
  24 #include <linux/slab.h>
  25 #include <linux/init.h>
  26 #include <linux/blkdev.h>
  27 #include <linux/parser.h>
  28 #include <linux/buffer_head.h>
  29 #include <linux/exportfs.h>
  30 #include <linux/vfs.h>
  31 #include <linux/random.h>
  32 #include <linux/mount.h>
  33 #include <linux/namei.h>
  34 #include <linux/quotaops.h>
  35 #include <linux/seq_file.h>
  36 #include <linux/proc_fs.h>
  37 #include <linux/ctype.h>
  38 #include <linux/log2.h>
  39 #include <linux/crc16.h>
  40 #include <linux/cleancache.h>
  41 #include <asm/uaccess.h>
  42
  43 #include <linux/kthread.h>
  44 #include <linux/freezer.h>
  45
  46 #include "ext4.h"
  47 #include "ext4_extents.h"       /* Needed for trace points definition */
  48 #include "ext4_jbd2.h"
  49 #include "xattr.h"
  50 #include "acl.h"
  51 #include "mballoc.h"
  52
  53 #define CREATE_TRACE_POINTS
  54 #include <trace/events/ext4.h>
  55
  56 static struct proc_dir_entry *ext4_proc_root;
  57 static struct kset *ext4_kset;
  58 static struct ext4_lazy_init *ext4_li_info;
  59 static struct mutex ext4_li_mtx;
  60 static struct ext4_features *ext4_feat;
  61 static int ext4_mballoc_ready;
  62
  63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
  64                              unsigned long journal_devnum);
  65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
  66 static int ext4_commit_super(struct super_block *sb, int sync);
  67 static void ext4_mark_recovery_complete(struct super_block *sb,
  68                                         struct ext4_super_block *es);
  69 static void ext4_clear_journal_err(struct super_block *sb,
  70                                    struct ext4_super_block *es);
  71 static int ext4_sync_fs(struct super_block *sb, int wait);
  72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
  73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
  74 static int ext4_unfreeze(struct super_block *sb);
  75 static int ext4_freeze(struct super_block *sb);
  76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
  77                        const char *dev_name, void *data);
  78 static inline int ext2_feature_set_ok(struct super_block *sb);
  79 static inline int ext3_feature_set_ok(struct super_block *sb);
  80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
  81 static void ext4_destroy_lazyinit_thread(void);
  82 static void ext4_unregister_li_request(struct super_block *sb);
  83 static void ext4_clear_request_list(void);
  84 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
  85
  86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
  87 static struct file_system_type ext2_fs_type = {
  88         .owner          = THIS_MODULE,
  89         .name           = "ext2",
  90         .mount          = ext4_mount,
  91         .kill_sb        = kill_block_super,
  92         .fs_flags       = FS_REQUIRES_DEV,
  93 };
  94 MODULE_ALIAS_FS("ext2");
  95 MODULE_ALIAS("ext2");
  96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
  97 #else
  98 #define IS_EXT2_SB(sb) (0)
  99 #endif
 100
 101
 102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
 103 static struct file_system_type ext3_fs_type = {
 104         .owner          = THIS_MODULE,
 105         .name           = "ext3",
 106         .mount          = ext4_mount,
 107         .kill_sb        = kill_block_super,
 108         .fs_flags       = FS_REQUIRES_DEV,
 109 };
 110 MODULE_ALIAS_FS("ext3");
 111 MODULE_ALIAS("ext3");
 112 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
 113 #else
 114 #define IS_EXT3_SB(sb) (0)
 115 #endif
 116
 117 static int ext4_verify_csum_type(struct super_block *sb,
 118                                  struct ext4_super_block *es)
 119 {
 120         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
 121                                         EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
 122                 return 1;
 123
 124         return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
 125 }
 126
 127 static __le32 ext4_superblock_csum(struct super_block *sb,
 128                                    struct ext4_super_block *es)
 129 {
 130         struct ext4_sb_info *sbi = EXT4_SB(sb);
 131         int offset = offsetof(struct ext4_super_block, s_checksum);
 132         __u32 csum;
 133
 134         csum = ext4_chksum(sbi, ~0, (char *)es, offset);
 135
 136         return cpu_to_le32(csum);
 137 }
 138
 139 static int ext4_superblock_csum_verify(struct super_block *sb,
 140                                        struct ext4_super_block *es)
 141 {
 142         if (!ext4_has_metadata_csum(sb))
 143                 return 1;
 144
 145         return es->s_checksum == ext4_superblock_csum(sb, es);
 146 }
 147
 148 void ext4_superblock_csum_set(struct super_block *sb)
 149 {
 150         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 151
 152         if (!ext4_has_metadata_csum(sb))
 153                 return;
 154
 155         es->s_checksum = ext4_superblock_csum(sb, es);
 156 }
 157
 158 void *ext4_kvmalloc(size_t size, gfp_t flags)
 159 {
 160         void *ret;
 161
 162         ret = kmalloc(size, flags | __GFP_NOWARN);
 163         if (!ret)
 164                 ret = __vmalloc(size, flags, PAGE_KERNEL);
 165         return ret;
 166 }
 167
 168 void *ext4_kvzalloc(size_t size, gfp_t flags)
 169 {
 170         void *ret;
 171
 172         ret = kzalloc(size, flags | __GFP_NOWARN);
 173         if (!ret)
 174                 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
 175         return ret;
 176 }
 177
 178 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
 179                                struct ext4_group_desc *bg)
 180 {
 181         return le32_to_cpu(bg->bg_block_bitmap_lo) |
 182                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 183                  (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
 184 }
 185
 186 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
 187                                struct ext4_group_desc *bg)
 188 {
 189         return le32_to_cpu(bg->bg_inode_bitmap_lo) |
 190                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 191                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
 192 }
 193
 194 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
 195                               struct ext4_group_desc *bg)
 196 {
 197         return le32_to_cpu(bg->bg_inode_table_lo) |
 198                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 199                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
 200 }
 201
 202 __u32 ext4_free_group_clusters(struct super_block *sb,
 203                                struct ext4_group_desc *bg)
 204 {
 205         return le16_to_cpu(bg->bg_free_blocks_count_lo) |
 206                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 207                  (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
 208 }
 209
 210 __u32 ext4_free_inodes_count(struct super_block *sb,
 211                               struct ext4_group_desc *bg)
 212 {
 213         return le16_to_cpu(bg->bg_free_inodes_count_lo) |
 214                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 215                  (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
 216 }
 217
 218 __u32 ext4_used_dirs_count(struct super_block *sb,
 219                               struct ext4_group_desc *bg)
 220 {
 221         return le16_to_cpu(bg->bg_used_dirs_count_lo) |
 222                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 223                  (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
 224 }
 225
 226 __u32 ext4_itable_unused_count(struct super_block *sb,
 227                               struct ext4_group_desc *bg)
 228 {
 229         return le16_to_cpu(bg->bg_itable_unused_lo) |
 230                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 231                  (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
 232 }
 233
 234 void ext4_block_bitmap_set(struct super_block *sb,
 235                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
 236 {
 237         bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
 238         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 239                 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
 240 }
 241
 242 void ext4_inode_bitmap_set(struct super_block *sb,
 243                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
 244 {
 245         bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
 246         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 247                 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
 248 }
 249
 250 void ext4_inode_table_set(struct super_block *sb,
 251                           struct ext4_group_desc *bg, ext4_fsblk_t blk)
 252 {
 253         bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
 254         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 255                 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
 256 }
 257
 258 void ext4_free_group_clusters_set(struct super_block *sb,
 259                                   struct ext4_group_desc *bg, __u32 count)
 260 {
 261         bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
 262         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 263                 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
 264 }
 265
 266 void ext4_free_inodes_set(struct super_block *sb,
 267                           struct ext4_group_desc *bg, __u32 count)
 268 {
 269         bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
 270         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 271                 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
 272 }
 273
 274 void ext4_used_dirs_set(struct super_block *sb,
 275                           struct ext4_group_desc *bg, __u32 count)
 276 {
 277         bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
 278         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 279                 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
 280 }
 281
 282 void ext4_itable_unused_set(struct super_block *sb,
 283                           struct ext4_group_desc *bg, __u32 count)
 284 {
 285         bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
 286         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 287                 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
 288 }
 289
 290
 291 static void __save_error_info(struct super_block *sb, const char *func,
 292                             unsigned int line)
 293 {
 294         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 295
 296         EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
 297         es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
 298         es->s_last_error_time = cpu_to_le32(get_seconds());
 299         strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
 300         es->s_last_error_line = cpu_to_le32(line);
 301         if (!es->s_first_error_time) {
 302                 es->s_first_error_time = es->s_last_error_time;
 303                 strncpy(es->s_first_error_func, func,
 304                         sizeof(es->s_first_error_func));
 305                 es->s_first_error_line = cpu_to_le32(line);
 306                 es->s_first_error_ino = es->s_last_error_ino;
 307                 es->s_first_error_block = es->s_last_error_block;
 308         }
 309         /*
 310          * Start the daily error reporting function if it hasn't been
 311          * started already
 312          */
 313         if (!es->s_error_count)
 314                 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
 315         le32_add_cpu(&es->s_error_count, 1);
 316 }
 317
 318 static void save_error_info(struct super_block *sb, const char *func,
 319                             unsigned int line)
 320 {
 321         __save_error_info(sb, func, line);
 322         ext4_commit_super(sb, 1);
 323 }
 324
 325 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
 326 {
 327         struct super_block              *sb = journal->j_private;
 328         struct ext4_sb_info             *sbi = EXT4_SB(sb);
 329         int                             error = is_journal_aborted(journal);
 330         struct ext4_journal_cb_entry    *jce;
 331
 332         BUG_ON(txn->t_state == T_FINISHED);
 333         spin_lock(&sbi->s_md_lock);
 334         while (!list_empty(&txn->t_private_list)) {
 335                 jce = list_entry(txn->t_private_list.next,
 336                                  struct ext4_journal_cb_entry, jce_list);
 337                 list_del_init(&jce->jce_list);
 338                 spin_unlock(&sbi->s_md_lock);
 339                 jce->jce_func(sb, jce, error);
 340                 spin_lock(&sbi->s_md_lock);
 341         }
 342         spin_unlock(&sbi->s_md_lock);
 343 }
 344
 345 /* Deal with the reporting of failure conditions on a filesystem such as
 346  * inconsistencies detected or read IO failures.
 347  *
 348  * On ext2, we can store the error state of the filesystem in the
 349  * superblock.  That is not possible on ext4, because we may have other
 350  * write ordering constraints on the superblock which prevent us from
 351  * writing it out straight away; and given that the journal is about to
 352  * be aborted, we can't rely on the current, or future, transactions to
 353  * write out the superblock safely.
 354  *
 355  * We'll just use the jbd2_journal_abort() error code to record an error in
 356  * the journal instead.  On recovery, the journal will complain about
 357  * that error until we've noted it down and cleared it.
 358  */
 359
 360 static void ext4_handle_error(struct super_block *sb)
 361 {
 362         if (sb->s_flags & MS_RDONLY)
 363                 return;
 364
 365         if (!test_opt(sb, ERRORS_CONT)) {
 366                 journal_t *journal = EXT4_SB(sb)->s_journal;
 367
 368                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
 369                 if (journal)
 370                         jbd2_journal_abort(journal, -EIO);
 371         }
 372         if (test_opt(sb, ERRORS_RO)) {
 373                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
 374                 /*
 375                  * Make sure updated value of ->s_mount_flags will be visible
 376                  * before ->s_flags update
 377                  */
 378                 smp_wmb();
 379                 sb->s_flags |= MS_RDONLY;
 380         }
 381         if (test_opt(sb, ERRORS_PANIC))
 382                 panic("EXT4-fs (device %s): panic forced after error\n",
 383                         sb->s_id);
 384 }
 385
 386 #define ext4_error_ratelimit(sb)                                        \
 387                 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),     \
 388                              "EXT4-fs error")
 389
 390 void __ext4_error(struct super_block *sb, const char *function,
 391                   unsigned int line, const char *fmt, ...)
 392 {
 393         struct va_format vaf;
 394         va_list args;
 395
 396         if (ext4_error_ratelimit(sb)) {
 397                 va_start(args, fmt);
 398                 vaf.fmt = fmt;
 399                 vaf.va = &args;
 400                 printk(KERN_CRIT
 401                        "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
 402                        sb->s_id, function, line, current->comm, &vaf);
 403                 va_end(args);
 404         }
 405         save_error_info(sb, function, line);
 406         ext4_handle_error(sb);
 407 }
 408
 409 void __ext4_error_inode(struct inode *inode, const char *function,
 410                         unsigned int line, ext4_fsblk_t block,
 411                         const char *fmt, ...)
 412 {
 413         va_list args;
 414         struct va_format vaf;
 415         struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
 416
 417         es->s_last_error_ino = cpu_to_le32(inode->i_ino);
 418         es->s_last_error_block = cpu_to_le64(block);
 419         if (ext4_error_ratelimit(inode->i_sb)) {
 420                 va_start(args, fmt);
 421                 vaf.fmt = fmt;
 422                 vaf.va = &args;
 423                 if (block)
 424                         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
 425                                "inode #%lu: block %llu: comm %s: %pV\n",
 426                                inode->i_sb->s_id, function, line, inode->i_ino,
 427                                block, current->comm, &vaf);
 428                 else
 429                         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
 430                                "inode #%lu: comm %s: %pV\n",
 431                                inode->i_sb->s_id, function, line, inode->i_ino,
 432                                current->comm, &vaf);
 433                 va_end(args);
 434         }
 435         save_error_info(inode->i_sb, function, line);
 436         ext4_handle_error(inode->i_sb);
 437 }
 438
 439 void __ext4_error_file(struct file *file, const char *function,
 440                        unsigned int line, ext4_fsblk_t block,
 441                        const char *fmt, ...)
 442 {
 443         va_list args;
 444         struct va_format vaf;
 445         struct ext4_super_block *es;
 446         struct inode *inode = file_inode(file);
 447         char pathname[80], *path;
 448
 449         es = EXT4_SB(inode->i_sb)->s_es;
 450         es->s_last_error_ino = cpu_to_le32(inode->i_ino);
 451         if (ext4_error_ratelimit(inode->i_sb)) {
 452                 path = d_path(&(file->f_path), pathname, sizeof(pathname));
 453                 if (IS_ERR(path))
 454                         path = "(unknown)";
 455                 va_start(args, fmt);
 456                 vaf.fmt = fmt;
 457                 vaf.va = &args;
 458                 if (block)
 459                         printk(KERN_CRIT
 460                                "EXT4-fs error (device %s): %s:%d: inode #%lu: "
 461                                "block %llu: comm %s: path %s: %pV\n",
 462                                inode->i_sb->s_id, function, line, inode->i_ino,
 463                                block, current->comm, path, &vaf);
 464                 else
 465                         printk(KERN_CRIT
 466                                "EXT4-fs error (device %s): %s:%d: inode #%lu: "
 467                                "comm %s: path %s: %pV\n",
 468                                inode->i_sb->s_id, function, line, inode->i_ino,
 469                                current->comm, path, &vaf);
 470                 va_end(args);
 471         }
 472         save_error_info(inode->i_sb, function, line);
 473         ext4_handle_error(inode->i_sb);
 474 }
 475
 476 const char *ext4_decode_error(struct super_block *sb, int errno,
 477                               char nbuf[16])
 478 {
 479         char *errstr = NULL;
 480
 481         switch (errno) {
 482         case -EIO:
 483                 errstr = "IO failure";
 484                 break;
 485         case -ENOMEM:
 486                 errstr = "Out of memory";
 487                 break;
 488         case -EROFS:
 489                 if (!sb || (EXT4_SB(sb)->s_journal &&
 490                             EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
 491                         errstr = "Journal has aborted";
 492                 else
 493                         errstr = "Readonly filesystem";
 494                 break;
 495         default:
 496                 /* If the caller passed in an extra buffer for unknown
 497                  * errors, textualise them now.  Else we just return
 498                  * NULL. */
 499                 if (nbuf) {
 500                         /* Check for truncated error codes... */
 501                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
 502                                 errstr = nbuf;
 503                 }
 504                 break;
 505         }
 506
 507         return errstr;
 508 }
 509
 510 /* __ext4_std_error decodes expected errors from journaling functions
 511  * automatically and invokes the appropriate error response.  */
 512
 513 void __ext4_std_error(struct super_block *sb, const char *function,
 514                       unsigned int line, int errno)
 515 {
 516         char nbuf[16];
 517         const char *errstr;
 518
 519         /* Special case: if the error is EROFS, and we're not already
 520          * inside a transaction, then there's really no point in logging
 521          * an error. */
 522         if (errno == -EROFS && journal_current_handle() == NULL &&
 523             (sb->s_flags & MS_RDONLY))
 524                 return;
 525
 526         if (ext4_error_ratelimit(sb)) {
 527                 errstr = ext4_decode_error(sb, errno, nbuf);
 528                 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
 529                        sb->s_id, function, line, errstr);
 530         }
 531
 532         save_error_info(sb, function, line);
 533         ext4_handle_error(sb);
 534 }
 535
 536 /*
 537  * ext4_abort is a much stronger failure handler than ext4_error.  The
 538  * abort function may be used to deal with unrecoverable failures such
 539  * as journal IO errors or ENOMEM at a critical moment in log management.
 540  *
 541  * We unconditionally force the filesystem into an ABORT|READONLY state,
 542  * unless the error response on the fs has been set to panic in which
 543  * case we take the easy way out and panic immediately.
 544  */
 545
 546 void __ext4_abort(struct super_block *sb, const char *function,
 547                 unsigned int line, const char *fmt, ...)
 548 {
 549         va_list args;
 550
 551         save_error_info(sb, function, line);
 552         va_start(args, fmt);
 553         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
 554                function, line);
 555         vprintk(fmt, args);
 556         printk("\n");
 557         va_end(args);
 558
 559         if ((sb->s_flags & MS_RDONLY) == 0) {
 560                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
 561                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
 562                 /*
 563                  * Make sure updated value of ->s_mount_flags will be visible
 564                  * before ->s_flags update
 565                  */
 566                 smp_wmb();
 567                 sb->s_flags |= MS_RDONLY;
 568                 if (EXT4_SB(sb)->s_journal)
 569                         jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
 570                 save_error_info(sb, function, line);
 571         }
 572         if (test_opt(sb, ERRORS_PANIC))
 573                 panic("EXT4-fs panic from previous error\n");
 574 }
 575
 576 void __ext4_msg(struct super_block *sb,
 577                 const char *prefix, const char *fmt, ...)
 578 {
 579         struct va_format vaf;
 580         va_list args;
 581
 582         if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
 583                 return;
 584
 585         va_start(args, fmt);
 586         vaf.fmt = fmt;
 587         vaf.va = &args;
 588         printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
 589         va_end(args);
 590 }
 591
 592 void __ext4_warning(struct super_block *sb, const char *function,
 593                     unsigned int line, const char *fmt, ...)
 594 {
 595         struct va_format vaf;
 596         va_list args;
 597
 598         if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
 599                           "EXT4-fs warning"))
 600                 return;
 601
 602         va_start(args, fmt);
 603         vaf.fmt = fmt;
 604         vaf.va = &args;
 605         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
 606                sb->s_id, function, line, &vaf);
 607         va_end(args);
 608 }
 609
 610 void __ext4_grp_locked_error(const char *function, unsigned int line,
 611                              struct super_block *sb, ext4_group_t grp,
 612                              unsigned long ino, ext4_fsblk_t block,
 613                              const char *fmt, ...)
 614 __releases(bitlock)
 615 __acquires(bitlock)
 616 {
 617         struct va_format vaf;
 618         va_list args;
 619         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 620
 621         es->s_last_error_ino = cpu_to_le32(ino);
 622         es->s_last_error_block = cpu_to_le64(block);
 623         __save_error_info(sb, function, line);
 624
 625         if (ext4_error_ratelimit(sb)) {
 626                 va_start(args, fmt);
 627                 vaf.fmt = fmt;
 628                 vaf.va = &args;
 629                 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
 630                        sb->s_id, function, line, grp);
 631                 if (ino)
 632                         printk(KERN_CONT "inode %lu: ", ino);
 633                 if (block)
 634                         printk(KERN_CONT "block %llu:",
 635                                (unsigned long long) block);
 636                 printk(KERN_CONT "%pV\n", &vaf);
 637                 va_end(args);
 638         }
 639
 640         if (test_opt(sb, ERRORS_CONT)) {
 641                 ext4_commit_super(sb, 0);
 642                 return;
 643         }
 644
 645         ext4_unlock_group(sb, grp);
 646         ext4_handle_error(sb);
 647         /*
 648          * We only get here in the ERRORS_RO case; relocking the group
 649          * may be dangerous, but nothing bad will happen since the
 650          * filesystem will have already been marked read/only and the
 651          * journal has been aborted.  We return 1 as a hint to callers
 652          * who might what to use the return value from
 653          * ext4_grp_locked_error() to distinguish between the
 654          * ERRORS_CONT and ERRORS_RO case, and perhaps return more
 655          * aggressively from the ext4 function in question, with a
 656          * more appropriate error code.
 657          */
 658         ext4_lock_group(sb, grp);
 659         return;
 660 }
 661
 662 void ext4_update_dynamic_rev(struct super_block *sb)
 663 {
 664         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 665
 666         if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
 667                 return;
 668
 669         ext4_warning(sb,
 670                      "updating to rev %d because of new feature flag, "
 671                      "running e2fsck is recommended",
 672                      EXT4_DYNAMIC_REV);
 673
 674         es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
 675         es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
 676         es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
 677         /* leave es->s_feature_*compat flags alone */
 678         /* es->s_uuid will be set by e2fsck if empty */
 679
 680         /*
 681          * The rest of the superblock fields should be zero, and if not it
 682          * means they are likely already in use, so leave them alone.  We
 683          * can leave it up to e2fsck to clean up any inconsistencies there.
 684          */
 685 }
 686
 687 /*
 688  * Open the external journal device
 689  */
 690 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
 691 {
 692         struct block_device *bdev;
 693         char b[BDEVNAME_SIZE];
 694
 695         bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
 696         if (IS_ERR(bdev))
 697                 goto fail;
 698         return bdev;
 699
 700 fail:
 701         ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
 702                         __bdevname(dev, b), PTR_ERR(bdev));
 703         return NULL;
 704 }
 705
 706 /*
 707  * Release the journal device
 708  */
 709 static void ext4_blkdev_put(struct block_device *bdev)
 710 {
 711         blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
 712 }
 713
 714 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
 715 {
 716         struct block_device *bdev;
 717         bdev = sbi->journal_bdev;
 718         if (bdev) {
 719                 ext4_blkdev_put(bdev);
 720                 sbi->journal_bdev = NULL;
 721         }
 722 }
 723
 724 static inline struct inode *orphan_list_entry(struct list_head *l)
 725 {
 726         return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
 727 }
 728
 729 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
 730 {
 731         struct list_head *l;
 732
 733         ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
 734                  le32_to_cpu(sbi->s_es->s_last_orphan));
 735
 736         printk(KERN_ERR "sb_info orphan list:\n");
 737         list_for_each(l, &sbi->s_orphan) {
 738                 struct inode *inode = orphan_list_entry(l);
 739                 printk(KERN_ERR "  "
 740                        "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
 741                        inode->i_sb->s_id, inode->i_ino, inode,
 742                        inode->i_mode, inode->i_nlink,
 743                        NEXT_ORPHAN(inode));
 744         }
 745 }
 746
 747 static void ext4_put_super(struct super_block *sb)
 748 {
 749         struct ext4_sb_info *sbi = EXT4_SB(sb);
 750         struct ext4_super_block *es = sbi->s_es;
 751         int i, err;
 752
 753         ext4_unregister_li_request(sb);
 754         dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
 755
 756         flush_workqueue(sbi->rsv_conversion_wq);
 757         destroy_workqueue(sbi->rsv_conversion_wq);
 758
 759         if (sbi->s_journal) {
 760                 err = jbd2_journal_destroy(sbi->s_journal);
 761                 sbi->s_journal = NULL;
 762                 if (err < 0)
 763                         ext4_abort(sb, "Couldn't clean up the journal");
 764         }
 765
 766         ext4_es_unregister_shrinker(sbi);
 767         del_timer_sync(&sbi->s_err_report);
 768         ext4_release_system_zone(sb);
 769         ext4_mb_release(sb);
 770         ext4_ext_release(sb);
 771         ext4_xattr_put_super(sb);
 772
 773         if (!(sb->s_flags & MS_RDONLY)) {
 774                 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
 775                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 776         }
 777         if (!(sb->s_flags & MS_RDONLY))
 778                 ext4_commit_super(sb, 1);
 779
 780         if (sbi->s_proc) {
 781                 remove_proc_entry("options", sbi->s_proc);
 782                 remove_proc_entry(sb->s_id, ext4_proc_root);
 783         }
 784         kobject_del(&sbi->s_kobj);
 785
 786         for (i = 0; i < sbi->s_gdb_count; i++)
 787                 brelse(sbi->s_group_desc[i]);
 788         kvfree(sbi->s_group_desc);
 789         kvfree(sbi->s_flex_groups);
 790         percpu_counter_destroy(&sbi->s_freeclusters_counter);
 791         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 792         percpu_counter_destroy(&sbi->s_dirs_counter);
 793         percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
 794         brelse(sbi->s_sbh);
 795 #ifdef CONFIG_QUOTA
 796         for (i = 0; i < EXT4_MAXQUOTAS; i++)
 797                 kfree(sbi->s_qf_names[i]);
 798 #endif
 799
 800         /* Debugging code just in case the in-memory inode orphan list
 801          * isn't empty.  The on-disk one can be non-empty if we've
 802          * detected an error and taken the fs readonly, but the
 803          * in-memory list had better be clean by this point. */
 804         if (!list_empty(&sbi->s_orphan))
 805                 dump_orphan_list(sb, sbi);
 806         J_ASSERT(list_empty(&sbi->s_orphan));
 807
 808         invalidate_bdev(sb->s_bdev);
 809         if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
 810                 /*
 811                  * Invalidate the journal device's buffers.  We don't want them
 812                  * floating about in memory - the physical journal device may
 813                  * hotswapped, and it breaks the `ro-after' testing code.
 814                  */
 815                 sync_blockdev(sbi->journal_bdev);
 816                 invalidate_bdev(sbi->journal_bdev);
 817                 ext4_blkdev_remove(sbi);
 818         }
 819         if (sbi->s_mb_cache) {
 820                 ext4_xattr_destroy_cache(sbi->s_mb_cache);
 821                 sbi->s_mb_cache = NULL;
 822         }
 823         if (sbi->s_mmp_tsk)
 824                 kthread_stop(sbi->s_mmp_tsk);
 825         sb->s_fs_info = NULL;
 826         /*
 827          * Now that we are completely done shutting down the
 828          * superblock, we need to actually destroy the kobject.
 829          */
 830         kobject_put(&sbi->s_kobj);
 831         wait_for_completion(&sbi->s_kobj_unregister);
 832         if (sbi->s_chksum_driver)
 833                 crypto_free_shash(sbi->s_chksum_driver);
 834         kfree(sbi->s_blockgroup_lock);
 835         kfree(sbi);
 836 }
 837
 838 static struct kmem_cache *ext4_inode_cachep;
 839
 840 /*
 841  * Called inside transaction, so use GFP_NOFS
 842  */
 843 static struct inode *ext4_alloc_inode(struct super_block *sb)
 844 {
 845         struct ext4_inode_info *ei;
 846
 847         ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
 848         if (!ei)
 849                 return NULL;
 850
 851         ei->vfs_inode.i_version = 1;
 852         spin_lock_init(&ei->i_raw_lock);
 853         INIT_LIST_HEAD(&ei->i_prealloc_list);
 854         spin_lock_init(&ei->i_prealloc_lock);
 855         ext4_es_init_tree(&ei->i_es_tree);
 856         rwlock_init(&ei->i_es_lock);
 857         INIT_LIST_HEAD(&ei->i_es_list);
 858         ei->i_es_all_nr = 0;
 859         ei->i_es_shk_nr = 0;
 860         ei->i_es_shrink_lblk = 0;
 861         ei->i_reserved_data_blocks = 0;
 862         ei->i_reserved_meta_blocks = 0;
 863         ei->i_allocated_meta_blocks = 0;
 864         ei->i_da_metadata_calc_len = 0;
 865         ei->i_da_metadata_calc_last_lblock = 0;
 866         spin_lock_init(&(ei->i_block_reservation_lock));
 867 #ifdef CONFIG_QUOTA
 868         ei->i_reserved_quota = 0;
 869         memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
 870 #endif
 871         ei->jinode = NULL;
 872         INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
 873         spin_lock_init(&ei->i_completed_io_lock);
 874         ei->i_sync_tid = 0;
 875         ei->i_datasync_tid = 0;
 876         atomic_set(&ei->i_ioend_count, 0);
 877         atomic_set(&ei->i_unwritten, 0);
 878         INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
 879
 880         return &ei->vfs_inode;
 881 }
 882
 883 static int ext4_drop_inode(struct inode *inode)
 884 {
 885         int drop = generic_drop_inode(inode);
 886
 887         trace_ext4_drop_inode(inode, drop);
 888         return drop;
 889 }
 890
 891 static void ext4_i_callback(struct rcu_head *head)
 892 {
 893         struct inode *inode = container_of(head, struct inode, i_rcu);
 894         kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
 895 }
 896
 897 static void ext4_destroy_inode(struct inode *inode)
 898 {
 899         if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
 900                 ext4_msg(inode->i_sb, KERN_ERR,
 901                          "Inode %lu (%p): orphan list check failed!",
 902                          inode->i_ino, EXT4_I(inode));
 903                 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
 904                                 EXT4_I(inode), sizeof(struct ext4_inode_info),
 905                                 true);
 906                 dump_stack();
 907         }
 908         call_rcu(&inode->i_rcu, ext4_i_callback);
 909 }
 910
 911 static void init_once(void *foo)
 912 {
 913         struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
 914
 915         INIT_LIST_HEAD(&ei->i_orphan);
 916         init_rwsem(&ei->xattr_sem);
 917         init_rwsem(&ei->i_data_sem);
 918         inode_init_once(&ei->vfs_inode);
 919 }
 920
 921 static int __init init_inodecache(void)
 922 {
 923         ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
 924                                              sizeof(struct ext4_inode_info),
 925                                              0, (SLAB_RECLAIM_ACCOUNT|
 926                                                 SLAB_MEM_SPREAD),
 927                                              init_once);
 928         if (ext4_inode_cachep == NULL)
 929                 return -ENOMEM;
 930         return 0;
 931 }
 932
 933 static void destroy_inodecache(void)
 934 {
 935         /*
 936          * Make sure all delayed rcu free inodes are flushed before we
 937          * destroy cache.
 938          */
 939         rcu_barrier();
 940         kmem_cache_destroy(ext4_inode_cachep);
 941 }
 942
 943 void ext4_clear_inode(struct inode *inode)
 944 {
 945         invalidate_inode_buffers(inode);
 946         clear_inode(inode);
 947         dquot_drop(inode);
 948         ext4_discard_preallocations(inode);
 949         ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
 950         if (EXT4_I(inode)->jinode) {
 951                 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
 952                                                EXT4_I(inode)->jinode);
 953                 jbd2_free_inode(EXT4_I(inode)->jinode);
 954                 EXT4_I(inode)->jinode = NULL;
 955         }
 956 }
 957
 958 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
 959                                         u64 ino, u32 generation)
 960 {
 961         struct inode *inode;
 962
 963         if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
 964                 return ERR_PTR(-ESTALE);
 965         if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
 966                 return ERR_PTR(-ESTALE);
 967
 968         /* iget isn't really right if the inode is currently unallocated!!
 969          *
 970          * ext4_read_inode will return a bad_inode if the inode had been
 971          * deleted, so we should be safe.
 972          *
 973          * Currently we don't know the generation for parent directory, so
 974          * a generation of 0 means "accept any"
 975          */
 976         inode = ext4_iget_normal(sb, ino);
 977         if (IS_ERR(inode))
 978                 return ERR_CAST(inode);
 979         if (generation && inode->i_generation != generation) {
 980                 iput(inode);
 981                 return ERR_PTR(-ESTALE);
 982         }
 983
 984         return inode;
 985 }
 986
 987 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
 988                                         int fh_len, int fh_type)
 989 {
 990         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 991                                     ext4_nfs_get_inode);
 992 }
 993
 994 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
 995                                         int fh_len, int fh_type)
 996 {
 997         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 998                                     ext4_nfs_get_inode);
 999 }
1000
1001 /*
1002  * Try to release metadata pages (indirect blocks, directories) which are
1003  * mapped via the block device.  Since these pages could have journal heads
1004  * which would prevent try_to_free_buffers() from freeing them, we must use
1005  * jbd2 layer's try_to_free_buffers() function to release them.
1006  */
1007 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1008                                  gfp_t wait)
1009 {
1010         journal_t *journal = EXT4_SB(sb)->s_journal;
1011
1012         WARN_ON(PageChecked(page));
1013         if (!page_has_buffers(page))
1014                 return 0;
1015         if (journal)
1016                 return jbd2_journal_try_to_free_buffers(journal, page,
1017                                                         wait & ~__GFP_WAIT);
1018         return try_to_free_buffers(page);
1019 }
1020
1021 #ifdef CONFIG_QUOTA
1022 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1023 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1024
1025 static int ext4_write_dquot(struct dquot *dquot);
1026 static int ext4_acquire_dquot(struct dquot *dquot);
1027 static int ext4_release_dquot(struct dquot *dquot);
1028 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1029 static int ext4_write_info(struct super_block *sb, int type);
1030 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1031                          struct path *path);
1032 static int ext4_quota_off(struct super_block *sb, int type);
1033 static int ext4_quota_on_mount(struct super_block *sb, int type);
1034 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1035                                size_t len, loff_t off);
1036 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1037                                 const char *data, size_t len, loff_t off);
1038 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1039                              unsigned int flags);
1040 static int ext4_enable_quotas(struct super_block *sb);
1041
1042 static struct dquot **ext4_get_dquots(struct inode *inode)
1043 {
1044         return EXT4_I(inode)->i_dquot;
1045 }
1046
1047 static const struct dquot_operations ext4_quota_operations = {
1048         .get_reserved_space = ext4_get_reserved_space,
1049         .write_dquot    = ext4_write_dquot,
1050         .acquire_dquot  = ext4_acquire_dquot,
1051         .release_dquot  = ext4_release_dquot,
1052         .mark_dirty     = ext4_mark_dquot_dirty,
1053         .write_info     = ext4_write_info,
1054         .alloc_dquot    = dquot_alloc,
1055         .destroy_dquot  = dquot_destroy,
1056 };
1057
1058 static const struct quotactl_ops ext4_qctl_operations = {
1059         .quota_on       = ext4_quota_on,
1060         .quota_off      = ext4_quota_off,
1061         .quota_sync     = dquot_quota_sync,
1062         .get_info       = dquot_get_dqinfo,
1063         .set_info       = dquot_set_dqinfo,
1064         .get_dqblk      = dquot_get_dqblk,
1065         .set_dqblk      = dquot_set_dqblk
1066 };
1067 #endif
1068
1069 static const struct super_operations ext4_sops = {
1070         .alloc_inode    = ext4_alloc_inode,
1071         .destroy_inode  = ext4_destroy_inode,
1072         .write_inode    = ext4_write_inode,
1073         .dirty_inode    = ext4_dirty_inode,
1074         .drop_inode     = ext4_drop_inode,
1075         .evict_inode    = ext4_evict_inode,
1076         .put_super      = ext4_put_super,
1077         .sync_fs        = ext4_sync_fs,
1078         .freeze_fs      = ext4_freeze,
1079         .unfreeze_fs    = ext4_unfreeze,
1080         .statfs         = ext4_statfs,
1081         .remount_fs     = ext4_remount,
1082         .show_options   = ext4_show_options,
1083 #ifdef CONFIG_QUOTA
1084         .quota_read     = ext4_quota_read,
1085         .quota_write    = ext4_quota_write,
1086         .get_dquots     = ext4_get_dquots,
1087 #endif
1088         .bdev_try_to_free_page = bdev_try_to_free_page,
1089 };
1090
1091 static const struct export_operations ext4_export_ops = {
1092         .fh_to_dentry = ext4_fh_to_dentry,
1093         .fh_to_parent = ext4_fh_to_parent,
1094         .get_parent = ext4_get_parent,
1095 };
1096
1097 enum {
1098         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1099         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1100         Opt_nouid32, Opt_debug, Opt_removed,
1101         Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1102         Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1103         Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1104         Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1105         Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1106         Opt_data_err_abort, Opt_data_err_ignore,
1107         Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1108         Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1109         Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1110         Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1111         Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1112         Opt_lazytime, Opt_nolazytime,
1113         Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1114         Opt_inode_readahead_blks, Opt_journal_ioprio,
1115         Opt_dioread_nolock, Opt_dioread_lock,
1116         Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1117         Opt_max_dir_size_kb, Opt_nojournal_checksum,
1118 };
1119
1120 static const match_table_t tokens = {
1121         {Opt_bsd_df, "bsddf"},
1122         {Opt_minix_df, "minixdf"},
1123         {Opt_grpid, "grpid"},
1124         {Opt_grpid, "bsdgroups"},
1125         {Opt_nogrpid, "nogrpid"},
1126         {Opt_nogrpid, "sysvgroups"},
1127         {Opt_resgid, "resgid=%u"},
1128         {Opt_resuid, "resuid=%u"},
1129         {Opt_sb, "sb=%u"},
1130         {Opt_err_cont, "errors=continue"},
1131         {Opt_err_panic, "errors=panic"},
1132         {Opt_err_ro, "errors=remount-ro"},
1133         {Opt_nouid32, "nouid32"},
1134         {Opt_debug, "debug"},
1135         {Opt_removed, "oldalloc"},
1136         {Opt_removed, "orlov"},
1137         {Opt_user_xattr, "user_xattr"},
1138         {Opt_nouser_xattr, "nouser_xattr"},
1139         {Opt_acl, "acl"},
1140         {Opt_noacl, "noacl"},
1141         {Opt_noload, "norecovery"},
1142         {Opt_noload, "noload"},
1143         {Opt_removed, "nobh"},
1144         {Opt_removed, "bh"},
1145         {Opt_commit, "commit=%u"},
1146         {Opt_min_batch_time, "min_batch_time=%u"},
1147         {Opt_max_batch_time, "max_batch_time=%u"},
1148         {Opt_journal_dev, "journal_dev=%u"},
1149         {Opt_journal_path, "journal_path=%s"},
1150         {Opt_journal_checksum, "journal_checksum"},
1151         {Opt_nojournal_checksum, "nojournal_checksum"},
1152         {Opt_journal_async_commit, "journal_async_commit"},
1153         {Opt_abort, "abort"},
1154         {Opt_data_journal, "data=journal"},
1155         {Opt_data_ordered, "data=ordered"},
1156         {Opt_data_writeback, "data=writeback"},
1157         {Opt_data_err_abort, "data_err=abort"},
1158         {Opt_data_err_ignore, "data_err=ignore"},
1159         {Opt_offusrjquota, "usrjquota="},
1160         {Opt_usrjquota, "usrjquota=%s"},
1161         {Opt_offgrpjquota, "grpjquota="},
1162         {Opt_grpjquota, "grpjquota=%s"},
1163         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1164         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1165         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1166         {Opt_grpquota, "grpquota"},
1167         {Opt_noquota, "noquota"},
1168         {Opt_quota, "quota"},
1169         {Opt_usrquota, "usrquota"},
1170         {Opt_barrier, "barrier=%u"},
1171         {Opt_barrier, "barrier"},
1172         {Opt_nobarrier, "nobarrier"},
1173         {Opt_i_version, "i_version"},
1174         {Opt_dax, "dax"},
1175         {Opt_stripe, "stripe=%u"},
1176         {Opt_delalloc, "delalloc"},
1177         {Opt_lazytime, "lazytime"},
1178         {Opt_nolazytime, "nolazytime"},
1179         {Opt_nodelalloc, "nodelalloc"},
1180         {Opt_removed, "mblk_io_submit"},
1181         {Opt_removed, "nomblk_io_submit"},
1182         {Opt_block_validity, "block_validity"},
1183         {Opt_noblock_validity, "noblock_validity"},
1184         {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1185         {Opt_journal_ioprio, "journal_ioprio=%u"},
1186         {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1187         {Opt_auto_da_alloc, "auto_da_alloc"},
1188         {Opt_noauto_da_alloc, "noauto_da_alloc"},
1189         {Opt_dioread_nolock, "dioread_nolock"},
1190         {Opt_dioread_lock, "dioread_lock"},
1191         {Opt_discard, "discard"},
1192         {Opt_nodiscard, "nodiscard"},
1193         {Opt_init_itable, "init_itable=%u"},
1194         {Opt_init_itable, "init_itable"},
1195         {Opt_noinit_itable, "noinit_itable"},
1196         {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1197         {Opt_removed, "check=none"},    /* mount option from ext2/3 */
1198         {Opt_removed, "nocheck"},       /* mount option from ext2/3 */
1199         {Opt_removed, "reservation"},   /* mount option from ext2/3 */
1200         {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1201         {Opt_removed, "journal=%u"},    /* mount option from ext2/3 */
1202         {Opt_err, NULL},
1203 };
1204
1205 static ext4_fsblk_t get_sb_block(void **data)
1206 {
1207         ext4_fsblk_t    sb_block;
1208         char            *options = (char *) *data;
1209
1210         if (!options || strncmp(options, "sb=", 3) != 0)
1211                 return 1;       /* Default location */
1212
1213         options += 3;
1214         /* TODO: use simple_strtoll with >32bit ext4 */
1215         sb_block = simple_strtoul(options, &options, 0);
1216         if (*options && *options != ',') {
1217                 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1218                        (char *) *data);
1219                 return 1;
1220         }
1221         if (*options == ',')
1222                 options++;
1223         *data = (void *) options;
1224
1225         return sb_block;
1226 }
1227
1228 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1229 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1230         "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1231
1232 #ifdef CONFIG_QUOTA
1233 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1234 {
1235         struct ext4_sb_info *sbi = EXT4_SB(sb);
1236         char *qname;
1237         int ret = -1;
1238
1239         if (sb_any_quota_loaded(sb) &&
1240                 !sbi->s_qf_names[qtype]) {
1241                 ext4_msg(sb, KERN_ERR,
1242                         "Cannot change journaled "
1243                         "quota options when quota turned on");
1244                 return -1;
1245         }
1246         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1247                 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1248                          "when QUOTA feature is enabled");
1249                 return -1;
1250         }
1251         qname = match_strdup(args);
1252         if (!qname) {
1253                 ext4_msg(sb, KERN_ERR,
1254                         "Not enough memory for storing quotafile name");
1255                 return -1;
1256         }
1257         if (sbi->s_qf_names[qtype]) {
1258                 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1259                         ret = 1;
1260                 else
1261                         ext4_msg(sb, KERN_ERR,
1262                                  "%s quota file already specified",
1263                                  QTYPE2NAME(qtype));
1264                 goto errout;
1265         }
1266         if (strchr(qname, '/')) {
1267                 ext4_msg(sb, KERN_ERR,
1268                         "quotafile must be on filesystem root");
1269                 goto errout;
1270         }
1271         sbi->s_qf_names[qtype] = qname;
1272         set_opt(sb, QUOTA);
1273         return 1;
1274 errout:
1275         kfree(qname);
1276         return ret;
1277 }
1278
1279 static int clear_qf_name(struct super_block *sb, int qtype)
1280 {
1281
1282         struct ext4_sb_info *sbi = EXT4_SB(sb);
1283
1284         if (sb_any_quota_loaded(sb) &&
1285                 sbi->s_qf_names[qtype]) {
1286                 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1287                         " when quota turned on");
1288                 return -1;
1289         }
1290         kfree(sbi->s_qf_names[qtype]);
1291         sbi->s_qf_names[qtype] = NULL;
1292         return 1;
1293 }
1294 #endif
1295
1296 #define MOPT_SET        0x0001
1297 #define MOPT_CLEAR      0x0002
1298 #define MOPT_NOSUPPORT  0x0004
1299 #define MOPT_EXPLICIT   0x0008
1300 #define MOPT_CLEAR_ERR  0x0010
1301 #define MOPT_GTE0       0x0020
1302 #ifdef CONFIG_QUOTA
1303 #define MOPT_Q          0
1304 #define MOPT_QFMT       0x0040
1305 #else
1306 #define MOPT_Q          MOPT_NOSUPPORT
1307 #define MOPT_QFMT       MOPT_NOSUPPORT
1308 #endif
1309 #define MOPT_DATAJ      0x0080
1310 #define MOPT_NO_EXT2    0x0100
1311 #define MOPT_NO_EXT3    0x0200
1312 #define MOPT_EXT4_ONLY  (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1313 #define MOPT_STRING     0x0400
1314
1315 static const struct mount_opts {
1316         int     token;
1317         int     mount_opt;
1318         int     flags;
1319 } ext4_mount_opts[] = {
1320         {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1321         {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1322         {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1323         {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1324         {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1325         {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1326         {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1327          MOPT_EXT4_ONLY | MOPT_SET},
1328         {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1329          MOPT_EXT4_ONLY | MOPT_CLEAR},
1330         {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1331         {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1332         {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1333          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1334         {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1335          MOPT_EXT4_ONLY | MOPT_CLEAR},
1336         {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1337          MOPT_EXT4_ONLY | MOPT_CLEAR},
1338         {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1339          MOPT_EXT4_ONLY | MOPT_SET},
1340         {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1341                                     EXT4_MOUNT_JOURNAL_CHECKSUM),
1342          MOPT_EXT4_ONLY | MOPT_SET},
1343         {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1344         {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1345         {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1346         {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1347         {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1348          MOPT_NO_EXT2 | MOPT_SET},
1349         {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1350          MOPT_NO_EXT2 | MOPT_CLEAR},
1351         {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1352         {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1353         {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1354         {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1355         {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1356         {Opt_commit, 0, MOPT_GTE0},
1357         {Opt_max_batch_time, 0, MOPT_GTE0},
1358         {Opt_min_batch_time, 0, MOPT_GTE0},
1359         {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1360         {Opt_init_itable, 0, MOPT_GTE0},
1361         {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1362         {Opt_stripe, 0, MOPT_GTE0},
1363         {Opt_resuid, 0, MOPT_GTE0},
1364         {Opt_resgid, 0, MOPT_GTE0},
1365         {Opt_journal_dev, 0, MOPT_GTE0},
1366         {Opt_journal_path, 0, MOPT_STRING},
1367         {Opt_journal_ioprio, 0, MOPT_GTE0},
1368         {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1369         {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1370         {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1371          MOPT_NO_EXT2 | MOPT_DATAJ},
1372         {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1373         {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1374 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1375         {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1376         {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1377 #else
1378         {Opt_acl, 0, MOPT_NOSUPPORT},
1379         {Opt_noacl, 0, MOPT_NOSUPPORT},
1380 #endif
1381         {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1382         {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1383         {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1384         {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1385                                                         MOPT_SET | MOPT_Q},
1386         {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1387                                                         MOPT_SET | MOPT_Q},
1388         {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1389                        EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1390         {Opt_usrjquota, 0, MOPT_Q},
1391         {Opt_grpjquota, 0, MOPT_Q},
1392         {Opt_offusrjquota, 0, MOPT_Q},
1393         {Opt_offgrpjquota, 0, MOPT_Q},
1394         {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1395         {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1396         {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1397         {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1398         {Opt_err, 0, 0}
1399 };
1400
1401 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1402                             substring_t *args, unsigned long *journal_devnum,
1403                             unsigned int *journal_ioprio, int is_remount)
1404 {
1405         struct ext4_sb_info *sbi = EXT4_SB(sb);
1406         const struct mount_opts *m;
1407         kuid_t uid;
1408         kgid_t gid;
1409         int arg = 0;
1410
1411 #ifdef CONFIG_QUOTA
1412         if (token == Opt_usrjquota)
1413                 return set_qf_name(sb, USRQUOTA, &args[0]);
1414         else if (token == Opt_grpjquota)
1415                 return set_qf_name(sb, GRPQUOTA, &args[0]);
1416         else if (token == Opt_offusrjquota)
1417                 return clear_qf_name(sb, USRQUOTA);
1418         else if (token == Opt_offgrpjquota)
1419                 return clear_qf_name(sb, GRPQUOTA);
1420 #endif
1421         switch (token) {
1422         case Opt_noacl:
1423         case Opt_nouser_xattr:
1424                 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1425                 break;
1426         case Opt_sb:
1427                 return 1;       /* handled by get_sb_block() */
1428         case Opt_removed:
1429                 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1430                 return 1;
1431         case Opt_abort:
1432                 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1433                 return 1;
1434         case Opt_i_version:
1435                 sb->s_flags |= MS_I_VERSION;
1436                 return 1;
1437         case Opt_lazytime:
1438                 sb->s_flags |= MS_LAZYTIME;
1439                 return 1;
1440         case Opt_nolazytime:
1441                 sb->s_flags &= ~MS_LAZYTIME;
1442                 return 1;
1443         }
1444
1445         for (m = ext4_mount_opts; m->token != Opt_err; m++)
1446                 if (token == m->token)
1447                         break;
1448
1449         if (m->token == Opt_err) {
1450                 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1451                          "or missing value", opt);
1452                 return -1;
1453         }
1454
1455         if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1456                 ext4_msg(sb, KERN_ERR,
1457                          "Mount option \"%s\" incompatible with ext2", opt);
1458                 return -1;
1459         }
1460         if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1461                 ext4_msg(sb, KERN_ERR,
1462                          "Mount option \"%s\" incompatible with ext3", opt);
1463                 return -1;
1464         }
1465
1466         if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1467                 return -1;
1468         if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1469                 return -1;
1470         if (m->flags & MOPT_EXPLICIT)
1471                 set_opt2(sb, EXPLICIT_DELALLOC);
1472         if (m->flags & MOPT_CLEAR_ERR)
1473                 clear_opt(sb, ERRORS_MASK);
1474         if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1475                 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1476                          "options when quota turned on");
1477                 return -1;
1478         }
1479
1480         if (m->flags & MOPT_NOSUPPORT) {
1481                 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1482         } else if (token == Opt_commit) {
1483                 if (arg == 0)
1484                         arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1485                 sbi->s_commit_interval = HZ * arg;
1486         } else if (token == Opt_max_batch_time) {
1487                 sbi->s_max_batch_time = arg;
1488         } else if (token == Opt_min_batch_time) {
1489                 sbi->s_min_batch_time = arg;
1490         } else if (token == Opt_inode_readahead_blks) {
1491                 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1492                         ext4_msg(sb, KERN_ERR,
1493                                  "EXT4-fs: inode_readahead_blks must be "
1494                                  "0 or a power of 2 smaller than 2^31");
1495                         return -1;
1496                 }
1497                 sbi->s_inode_readahead_blks = arg;
1498         } else if (token == Opt_init_itable) {
1499                 set_opt(sb, INIT_INODE_TABLE);
1500                 if (!args->from)
1501                         arg = EXT4_DEF_LI_WAIT_MULT;
1502                 sbi->s_li_wait_mult = arg;
1503         } else if (token == Opt_max_dir_size_kb) {
1504                 sbi->s_max_dir_size_kb = arg;
1505         } else if (token == Opt_stripe) {
1506                 sbi->s_stripe = arg;
1507         } else if (token == Opt_resuid) {
1508                 uid = make_kuid(current_user_ns(), arg);
1509                 if (!uid_valid(uid)) {
1510                         ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1511                         return -1;
1512                 }
1513                 sbi->s_resuid = uid;
1514         } else if (token == Opt_resgid) {
1515                 gid = make_kgid(current_user_ns(), arg);
1516                 if (!gid_valid(gid)) {
1517                         ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1518                         return -1;
1519                 }
1520                 sbi->s_resgid = gid;
1521         } else if (token == Opt_journal_dev) {
1522                 if (is_remount) {
1523                         ext4_msg(sb, KERN_ERR,
1524                                  "Cannot specify journal on remount");
1525                         return -1;
1526                 }
1527                 *journal_devnum = arg;
1528         } else if (token == Opt_journal_path) {
1529                 char *journal_path;
1530                 struct inode *journal_inode;
1531                 struct path path;
1532                 int error;
1533
1534                 if (is_remount) {
1535                         ext4_msg(sb, KERN_ERR,
1536                                  "Cannot specify journal on remount");
1537                         return -1;
1538                 }
1539                 journal_path = match_strdup(&args[0]);
1540                 if (!journal_path) {
1541                         ext4_msg(sb, KERN_ERR, "error: could not dup "
1542                                 "journal device string");
1543                         return -1;
1544                 }
1545
1546                 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1547                 if (error) {
1548                         ext4_msg(sb, KERN_ERR, "error: could not find "
1549                                 "journal device path: error %d", error);
1550                         kfree(journal_path);
1551                         return -1;
1552                 }
1553
1554                 journal_inode = path.dentry->d_inode;
1555                 if (!S_ISBLK(journal_inode->i_mode)) {
1556                         ext4_msg(sb, KERN_ERR, "error: journal path %s "
1557                                 "is not a block device", journal_path);
1558                         path_put(&path);
1559                         kfree(journal_path);
1560                         return -1;
1561                 }
1562
1563                 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1564                 path_put(&path);
1565                 kfree(journal_path);
1566         } else if (token == Opt_journal_ioprio) {
1567                 if (arg > 7) {
1568                         ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1569                                  " (must be 0-7)");
1570                         return -1;
1571                 }
1572                 *journal_ioprio =
1573                         IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1574         } else if (m->flags & MOPT_DATAJ) {
1575                 if (is_remount) {
1576                         if (!sbi->s_journal)
1577                                 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1578                         else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1579                                 ext4_msg(sb, KERN_ERR,
1580                                          "Cannot change data mode on remount");
1581                                 return -1;
1582                         }
1583                 } else {
1584                         clear_opt(sb, DATA_FLAGS);
1585                         sbi->s_mount_opt |= m->mount_opt;
1586                 }
1587 #ifdef CONFIG_QUOTA
1588         } else if (m->flags & MOPT_QFMT) {
1589                 if (sb_any_quota_loaded(sb) &&
1590                     sbi->s_jquota_fmt != m->mount_opt) {
1591                         ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1592                                  "quota options when quota turned on");
1593                         return -1;
1594                 }
1595                 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1596                                                EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1597                         ext4_msg(sb, KERN_ERR,
1598                                  "Cannot set journaled quota options "
1599                                  "when QUOTA feature is enabled");
1600                         return -1;
1601                 }
1602                 sbi->s_jquota_fmt = m->mount_opt;
1603 #endif
1604 #ifndef CONFIG_FS_DAX
1605         } else if (token == Opt_dax) {
1606                 ext4_msg(sb, KERN_INFO, "dax option not supported");
1607                 return -1;
1608 #endif
1609         } else {
1610                 if (!args->from)
1611                         arg = 1;
1612                 if (m->flags & MOPT_CLEAR)
1613                         arg = !arg;
1614                 else if (unlikely(!(m->flags & MOPT_SET))) {
1615                         ext4_msg(sb, KERN_WARNING,
1616                                  "buggy handling of option %s", opt);
1617                         WARN_ON(1);
1618                         return -1;
1619                 }
1620                 if (arg != 0)
1621                         sbi->s_mount_opt |= m->mount_opt;
1622                 else
1623                         sbi->s_mount_opt &= ~m->mount_opt;
1624         }
1625         return 1;
1626 }
1627
1628 static int parse_options(char *options, struct super_block *sb,
1629                          unsigned long *journal_devnum,
1630                          unsigned int *journal_ioprio,
1631                          int is_remount)
1632 {
1633         struct ext4_sb_info *sbi = EXT4_SB(sb);
1634         char *p;
1635         substring_t args[MAX_OPT_ARGS];
1636         int token;
1637
1638         if (!options)
1639                 return 1;
1640
1641         while ((p = strsep(&options, ",")) != NULL) {
1642                 if (!*p)
1643                         continue;
1644                 /*
1645                  * Initialize args struct so we know whether arg was
1646                  * found; some options take optional arguments.
1647                  */
1648                 args[0].to = args[0].from = NULL;
1649                 token = match_token(p, tokens, args);
1650                 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1651                                      journal_ioprio, is_remount) < 0)
1652                         return 0;
1653         }
1654 #ifdef CONFIG_QUOTA
1655         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1656             (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1657                 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1658                          "feature is enabled");
1659                 return 0;
1660         }
1661         if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1662                 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1663                         clear_opt(sb, USRQUOTA);
1664
1665                 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1666                         clear_opt(sb, GRPQUOTA);
1667
1668                 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1669                         ext4_msg(sb, KERN_ERR, "old and new quota "
1670                                         "format mixing");
1671                         return 0;
1672                 }
1673
1674                 if (!sbi->s_jquota_fmt) {
1675                         ext4_msg(sb, KERN_ERR, "journaled quota format "
1676                                         "not specified");
1677                         return 0;
1678                 }
1679         }
1680 #endif
1681         if (test_opt(sb, DIOREAD_NOLOCK)) {
1682                 int blocksize =
1683                         BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1684
1685                 if (blocksize < PAGE_CACHE_SIZE) {
1686                         ext4_msg(sb, KERN_ERR, "can't mount with "
1687                                  "dioread_nolock if block size != PAGE_SIZE");
1688                         return 0;
1689                 }
1690         }
1691         if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1692             test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1693                 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1694                          "in data=ordered mode");
1695                 return 0;
1696         }
1697         return 1;
1698 }
1699
1700 static inline void ext4_show_quota_options(struct seq_file *seq,
1701                                            struct super_block *sb)
1702 {
1703 #if defined(CONFIG_QUOTA)
1704         struct ext4_sb_info *sbi = EXT4_SB(sb);
1705
1706         if (sbi->s_jquota_fmt) {
1707                 char *fmtname = "";
1708
1709                 switch (sbi->s_jquota_fmt) {
1710                 case QFMT_VFS_OLD:
1711                         fmtname = "vfsold";
1712                         break;
1713                 case QFMT_VFS_V0:
1714                         fmtname = "vfsv0";
1715                         break;
1716                 case QFMT_VFS_V1:
1717                         fmtname = "vfsv1";
1718                         break;
1719                 }
1720                 seq_printf(seq, ",jqfmt=%s", fmtname);
1721         }
1722
1723         if (sbi->s_qf_names[USRQUOTA])
1724                 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1725
1726         if (sbi->s_qf_names[GRPQUOTA])
1727                 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1728 #endif
1729 }
1730
1731 static const char *token2str(int token)
1732 {
1733         const struct match_token *t;
1734
1735         for (t = tokens; t->token != Opt_err; t++)
1736                 if (t->token == token && !strchr(t->pattern, '='))
1737                         break;
1738         return t->pattern;
1739 }
1740
1741 /*
1742  * Show an option if
1743  *  - it's set to a non-default value OR
1744  *  - if the per-sb default is different from the global default
1745  */
1746 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1747                               int nodefs)
1748 {
1749         struct ext4_sb_info *sbi = EXT4_SB(sb);
1750         struct ext4_super_block *es = sbi->s_es;
1751         int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1752         const struct mount_opts *m;
1753         char sep = nodefs ? '\n' : ',';
1754
1755 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1756 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1757
1758         if (sbi->s_sb_block != 1)
1759                 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1760
1761         for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1762                 int want_set = m->flags & MOPT_SET;
1763                 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1764                     (m->flags & MOPT_CLEAR_ERR))
1765                         continue;
1766                 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1767                         continue; /* skip if same as the default */
1768                 if ((want_set &&
1769                      (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1770                     (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1771                         continue; /* select Opt_noFoo vs Opt_Foo */
1772                 SEQ_OPTS_PRINT("%s", token2str(m->token));
1773         }
1774
1775         if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1776             le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1777                 SEQ_OPTS_PRINT("resuid=%u",
1778                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1779         if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1780             le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1781                 SEQ_OPTS_PRINT("resgid=%u",
1782                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1783         def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1784         if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1785                 SEQ_OPTS_PUTS("errors=remount-ro");
1786         if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1787                 SEQ_OPTS_PUTS("errors=continue");
1788         if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1789                 SEQ_OPTS_PUTS("errors=panic");
1790         if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1791                 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1792         if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1793                 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1794         if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1795                 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1796         if (sb->s_flags & MS_I_VERSION)
1797                 SEQ_OPTS_PUTS("i_version");
1798         if (nodefs || sbi->s_stripe)
1799                 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1800         if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1801                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1802                         SEQ_OPTS_PUTS("data=journal");
1803                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1804                         SEQ_OPTS_PUTS("data=ordered");
1805                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1806                         SEQ_OPTS_PUTS("data=writeback");
1807         }
1808         if (nodefs ||
1809             sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1810                 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1811                                sbi->s_inode_readahead_blks);
1812
1813         if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1814                        (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1815                 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1816         if (nodefs || sbi->s_max_dir_size_kb)
1817                 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1818
1819         ext4_show_quota_options(seq, sb);
1820         return 0;
1821 }
1822
1823 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1824 {
1825         return _ext4_show_options(seq, root->d_sb, 0);
1826 }
1827
1828 static int options_seq_show(struct seq_file *seq, void *offset)
1829 {
1830         struct super_block *sb = seq->private;
1831         int rc;
1832
1833         seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1834         rc = _ext4_show_options(seq, sb, 1);
1835         seq_puts(seq, "\n");
1836         return rc;
1837 }
1838
1839 static int options_open_fs(struct inode *inode, struct file *file)
1840 {
1841         return single_open(file, options_seq_show, PDE_DATA(inode));
1842 }
1843
1844 static const struct file_operations ext4_seq_options_fops = {
1845         .owner = THIS_MODULE,
1846         .open = options_open_fs,
1847         .read = seq_read,
1848         .llseek = seq_lseek,
1849         .release = single_release,
1850 };
1851
1852 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1853                             int read_only)
1854 {
1855         struct ext4_sb_info *sbi = EXT4_SB(sb);
1856         int res = 0;
1857
1858         if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1859                 ext4_msg(sb, KERN_ERR, "revision level too high, "
1860                          "forcing read-only mode");
1861                 res = MS_RDONLY;
1862         }
1863         if (read_only)
1864                 goto done;
1865         if (!(sbi->s_mount_state & EXT4_VALID_FS))
1866                 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1867                          "running e2fsck is recommended");
1868         else if (sbi->s_mount_state & EXT4_ERROR_FS)
1869                 ext4_msg(sb, KERN_WARNING,
1870                          "warning: mounting fs with errors, "
1871                          "running e2fsck is recommended");
1872         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1873                  le16_to_cpu(es->s_mnt_count) >=
1874                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1875                 ext4_msg(sb, KERN_WARNING,
1876                          "warning: maximal mount count reached, "
1877                          "running e2fsck is recommended");
1878         else if (le32_to_cpu(es->s_checkinterval) &&
1879                 (le32_to_cpu(es->s_lastcheck) +
1880                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1881                 ext4_msg(sb, KERN_WARNING,
1882                          "warning: checktime reached, "
1883                          "running e2fsck is recommended");
1884         if (!sbi->s_journal)
1885                 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1886         if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1887                 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1888         le16_add_cpu(&es->s_mnt_count, 1);
1889         es->s_mtime = cpu_to_le32(get_seconds());
1890         ext4_update_dynamic_rev(sb);
1891         if (sbi->s_journal)
1892                 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1893
1894         ext4_commit_super(sb, 1);
1895 done:
1896         if (test_opt(sb, DEBUG))
1897                 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1898                                 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1899                         sb->s_blocksize,
1900                         sbi->s_groups_count,
1901                         EXT4_BLOCKS_PER_GROUP(sb),
1902                         EXT4_INODES_PER_GROUP(sb),
1903                         sbi->s_mount_opt, sbi->s_mount_opt2);
1904
1905         cleancache_init_fs(sb);
1906         return res;
1907 }
1908
1909 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1910 {
1911         struct ext4_sb_info *sbi = EXT4_SB(sb);
1912         struct flex_groups *new_groups;
1913         int size;
1914
1915         if (!sbi->s_log_groups_per_flex)
1916                 return 0;
1917
1918         size = ext4_flex_group(sbi, ngroup - 1) + 1;
1919         if (size <= sbi->s_flex_groups_allocated)
1920                 return 0;
1921
1922         size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1923         new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1924         if (!new_groups) {
1925                 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1926                          size / (int) sizeof(struct flex_groups));
1927                 return -ENOMEM;
1928         }
1929
1930         if (sbi->s_flex_groups) {
1931                 memcpy(new_groups, sbi->s_flex_groups,
1932                        (sbi->s_flex_groups_allocated *
1933                         sizeof(struct flex_groups)));
1934                 kvfree(sbi->s_flex_groups);
1935         }
1936         sbi->s_flex_groups = new_groups;
1937         sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1938         return 0;
1939 }
1940
1941 static int ext4_fill_flex_info(struct super_block *sb)
1942 {
1943         struct ext4_sb_info *sbi = EXT4_SB(sb);
1944         struct ext4_group_desc *gdp = NULL;
1945         ext4_group_t flex_group;
1946         int i, err;
1947
1948         sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1949         if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1950                 sbi->s_log_groups_per_flex = 0;
1951                 return 1;
1952         }
1953
1954         err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1955         if (err)
1956                 goto failed;
1957
1958         for (i = 0; i < sbi->s_groups_count; i++) {
1959                 gdp = ext4_get_group_desc(sb, i, NULL);
1960
1961                 flex_group = ext4_flex_group(sbi, i);
1962                 atomic_add(ext4_free_inodes_count(sb, gdp),
1963                            &sbi->s_flex_groups[flex_group].free_inodes);
1964                 atomic64_add(ext4_free_group_clusters(sb, gdp),
1965                              &sbi->s_flex_groups[flex_group].free_clusters);
1966                 atomic_add(ext4_used_dirs_count(sb, gdp),
1967                            &sbi->s_flex_groups[flex_group].used_dirs);
1968         }
1969
1970         return 1;
1971 failed:
1972         return 0;
1973 }
1974
1975 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1976                                    struct ext4_group_desc *gdp)
1977 {
1978         int offset;
1979         __u16 crc = 0;
1980         __le32 le_group = cpu_to_le32(block_group);
1981
1982         if (ext4_has_metadata_csum(sbi->s_sb)) {
1983                 /* Use new metadata_csum algorithm */
1984                 __le16 save_csum;
1985                 __u32 csum32;
1986
1987                 save_csum = gdp->bg_checksum;
1988                 gdp->bg_checksum = 0;
1989                 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1990                                      sizeof(le_group));
1991                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1992                                      sbi->s_desc_size);
1993                 gdp->bg_checksum = save_csum;
1994
1995                 crc = csum32 & 0xFFFF;
1996                 goto out;
1997         }
1998
1999         /* old crc16 code */
2000         if (!(sbi->s_es->s_feature_ro_compat &
2001               cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2002                 return 0;
2003
2004         offset = offsetof(struct ext4_group_desc, bg_checksum);
2005
2006         crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2007         crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2008         crc = crc16(crc, (__u8 *)gdp, offset);
2009         offset += sizeof(gdp->bg_checksum); /* skip checksum */
2010         /* for checksum of struct ext4_group_desc do the rest...*/
2011         if ((sbi->s_es->s_feature_incompat &
2012              cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2013             offset < le16_to_cpu(sbi->s_es->s_desc_size))
2014                 crc = crc16(crc, (__u8 *)gdp + offset,
2015                             le16_to_cpu(sbi->s_es->s_desc_size) -
2016                                 offset);
2017
2018 out:
2019         return cpu_to_le16(crc);
2020 }
2021
2022 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2023                                 struct ext4_group_desc *gdp)
2024 {
2025         if (ext4_has_group_desc_csum(sb) &&
2026             (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2027                                                       block_group, gdp)))
2028                 return 0;
2029
2030         return 1;
2031 }
2032
2033 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2034                               struct ext4_group_desc *gdp)
2035 {
2036         if (!ext4_has_group_desc_csum(sb))
2037                 return;
2038         gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2039 }
2040
2041 /* Called at mount-time, super-block is locked */
2042 static int ext4_check_descriptors(struct super_block *sb,
2043                                   ext4_group_t *first_not_zeroed)
2044 {
2045         struct ext4_sb_info *sbi = EXT4_SB(sb);
2046         ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2047         ext4_fsblk_t last_block;
2048         ext4_fsblk_t block_bitmap;
2049         ext4_fsblk_t inode_bitmap;
2050         ext4_fsblk_t inode_table;
2051         int flexbg_flag = 0;
2052         ext4_group_t i, grp = sbi->s_groups_count;
2053
2054         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2055                 flexbg_flag = 1;
2056
2057         ext4_debug("Checking group descriptors");
2058
2059         for (i = 0; i < sbi->s_groups_count; i++) {
2060                 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2061
2062                 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2063                         last_block = ext4_blocks_count(sbi->s_es) - 1;
2064                 else
2065                         last_block = first_block +
2066                                 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2067
2068                 if ((grp == sbi->s_groups_count) &&
2069                    !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2070                         grp = i;
2071
2072                 block_bitmap = ext4_block_bitmap(sb, gdp);
2073                 if (block_bitmap < first_block || block_bitmap > last_block) {
2074                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2075                                "Block bitmap for group %u not in group "
2076                                "(block %llu)!", i, block_bitmap);
2077                         return 0;
2078                 }
2079                 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2080                 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2081                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2082                                "Inode bitmap for group %u not in group "
2083                                "(block %llu)!", i, inode_bitmap);
2084                         return 0;
2085                 }
2086                 inode_table = ext4_inode_table(sb, gdp);
2087                 if (inode_table < first_block ||
2088                     inode_table + sbi->s_itb_per_group - 1 > last_block) {
2089                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2090                                "Inode table for group %u not in group "
2091                                "(block %llu)!", i, inode_table);
2092                         return 0;
2093                 }
2094                 ext4_lock_group(sb, i);
2095                 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2096                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2097                                  "Checksum for group %u failed (%u!=%u)",
2098                                  i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2099                                      gdp)), le16_to_cpu(gdp->bg_checksum));
2100                         if (!(sb->s_flags & MS_RDONLY)) {
2101                                 ext4_unlock_group(sb, i);
2102                                 return 0;
2103                         }
2104                 }
2105                 ext4_unlock_group(sb, i);
2106                 if (!flexbg_flag)
2107                         first_block += EXT4_BLOCKS_PER_GROUP(sb);
2108         }
2109         if (NULL != first_not_zeroed)
2110                 *first_not_zeroed = grp;
2111         return 1;
2112 }
2113
2114 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2115  * the superblock) which were deleted from all directories, but held open by
2116  * a process at the time of a crash.  We walk the list and try to delete these
2117  * inodes at recovery time (only with a read-write filesystem).
2118  *
2119  * In order to keep the orphan inode chain consistent during traversal (in
2120  * case of crash during recovery), we link each inode into the superblock
2121  * orphan list_head and handle it the same way as an inode deletion during
2122  * normal operation (which journals the operations for us).
2123  *
2124  * We only do an iget() and an iput() on each inode, which is very safe if we
2125  * accidentally point at an in-use or already deleted inode.  The worst that
2126  * can happen in this case is that we get a "bit already cleared" message from
2127  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2128  * e2fsck was run on this filesystem, and it must have already done the orphan
2129  * inode cleanup for us, so we can safely abort without any further action.
2130  */
2131 static void ext4_orphan_cleanup(struct super_block *sb,
2132                                 struct ext4_super_block *es)
2133 {
2134         unsigned int s_flags = sb->s_flags;
2135         int nr_orphans = 0, nr_truncates = 0;
2136 #ifdef CONFIG_QUOTA
2137         int i;
2138 #endif
2139         if (!es->s_last_orphan) {
2140                 jbd_debug(4, "no orphan inodes to clean up\n");
2141                 return;
2142         }
2143
2144         if (bdev_read_only(sb->s_bdev)) {
2145                 ext4_msg(sb, KERN_ERR, "write access "
2146                         "unavailable, skipping orphan cleanup");
2147                 return;
2148         }
2149
2150         /* Check if feature set would not allow a r/w mount */
2151         if (!ext4_feature_set_ok(sb, 0)) {
2152                 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2153                          "unknown ROCOMPAT features");
2154                 return;
2155         }
2156
2157         if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2158                 /* don't clear list on RO mount w/ errors */
2159                 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2160                         ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2161                                   "clearing orphan list.\n");
2162                         es->s_last_orphan = 0;
2163                 }
2164                 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2165                 return;
2166         }
2167
2168         if (s_flags & MS_RDONLY) {
2169                 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2170                 sb->s_flags &= ~MS_RDONLY;
2171         }
2172 #ifdef CONFIG_QUOTA
2173         /* Needed for iput() to work correctly and not trash data */
2174         sb->s_flags |= MS_ACTIVE;
2175         /* Turn on quotas so that they are updated correctly */
2176         for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2177                 if (EXT4_SB(sb)->s_qf_names[i]) {
2178                         int ret = ext4_quota_on_mount(sb, i);
2179                         if (ret < 0)
2180                                 ext4_msg(sb, KERN_ERR,
2181                                         "Cannot turn on journaled "
2182                                         "quota: error %d", ret);
2183                 }
2184         }
2185 #endif
2186
2187         while (es->s_last_orphan) {
2188                 struct inode *inode;
2189
2190                 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2191                 if (IS_ERR(inode)) {
2192                         es->s_last_orphan = 0;
2193                         break;
2194                 }
2195
2196                 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2197                 dquot_initialize(inode);
2198                 if (inode->i_nlink) {
2199                         if (test_opt(sb, DEBUG))
2200                                 ext4_msg(sb, KERN_DEBUG,
2201                                         "%s: truncating inode %lu to %lld bytes",
2202                                         __func__, inode->i_ino, inode->i_size);
2203                         jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2204                                   inode->i_ino, inode->i_size);
2205                         mutex_lock(&inode->i_mutex);
2206                         truncate_inode_pages(inode->i_mapping, inode->i_size);
2207                         ext4_truncate(inode);
2208                         mutex_unlock(&inode->i_mutex);
2209                         nr_truncates++;
2210                 } else {
2211                         if (test_opt(sb, DEBUG))
2212                                 ext4_msg(sb, KERN_DEBUG,
2213                                         "%s: deleting unreferenced inode %lu",
2214                                         __func__, inode->i_ino);
2215                         jbd_debug(2, "deleting unreferenced inode %lu\n",
2216                                   inode->i_ino);
2217                         nr_orphans++;
2218                 }
2219                 iput(inode);  /* The delete magic happens here! */
2220         }
2221
2222 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2223
2224         if (nr_orphans)
2225                 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2226                        PLURAL(nr_orphans));
2227         if (nr_truncates)
2228                 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2229                        PLURAL(nr_truncates));
2230 #ifdef CONFIG_QUOTA
2231         /* Turn quotas off */
2232         for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2233                 if (sb_dqopt(sb)->files[i])
2234                         dquot_quota_off(sb, i);
2235         }
2236 #endif
2237         sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2238 }
2239
2240 /*
2241  * Maximal extent format file size.
2242  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2243  * extent format containers, within a sector_t, and within i_blocks
2244  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2245  * so that won't be a limiting factor.
2246  *
2247  * However there is other limiting factor. We do store extents in the form
2248  * of starting block and length, hence the resulting length of the extent
2249  * covering maximum file size must fit into on-disk format containers as
2250  * well. Given that length is always by 1 unit bigger than max unit (because
2251  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2252  *
2253  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2254  */
2255 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2256 {
2257         loff_t res;
2258         loff_t upper_limit = MAX_LFS_FILESIZE;
2259
2260         /* small i_blocks in vfs inode? */
2261         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2262                 /*
2263                  * CONFIG_LBDAF is not enabled implies the inode
2264                  * i_block represent total blocks in 512 bytes
2265                  * 32 == size of vfs inode i_blocks * 8
2266                  */
2267                 upper_limit = (1LL << 32) - 1;
2268
2269                 /* total blocks in file system block size */
2270                 upper_limit >>= (blkbits - 9);
2271                 upper_limit <<= blkbits;
2272         }
2273
2274         /*
2275          * 32-bit extent-start container, ee_block. We lower the maxbytes
2276          * by one fs block, so ee_len can cover the extent of maximum file
2277          * size
2278          */
2279         res = (1LL << 32) - 1;
2280         res <<= blkbits;
2281
2282         /* Sanity check against vm- & vfs- imposed limits */
2283         if (res > upper_limit)
2284                 res = upper_limit;
2285
2286         return res;
2287 }
2288
2289 /*
2290  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2291  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2292  * We need to be 1 filesystem block less than the 2^48 sector limit.
2293  */
2294 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2295 {
2296         loff_t res = EXT4_NDIR_BLOCKS;
2297         int meta_blocks;
2298         loff_t upper_limit;
2299         /* This is calculated to be the largest file size for a dense, block
2300          * mapped file such that the file's total number of 512-byte sectors,
2301          * including data and all indirect blocks, does not exceed (2^48 - 1).
2302          *
2303          * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2304          * number of 512-byte sectors of the file.
2305          */
2306
2307         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2308                 /*
2309                  * !has_huge_files or CONFIG_LBDAF not enabled implies that
2310                  * the inode i_block field represents total file blocks in
2311                  * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2312                  */
2313                 upper_limit = (1LL << 32) - 1;
2314
2315                 /* total blocks in file system block size */
2316                 upper_limit >>= (bits - 9);
2317
2318         } else {
2319                 /*
2320                  * We use 48 bit ext4_inode i_blocks
2321                  * With EXT4_HUGE_FILE_FL set the i_blocks
2322                  * represent total number of blocks in
2323                  * file system block size
2324                  */
2325                 upper_limit = (1LL << 48) - 1;
2326
2327         }
2328
2329         /* indirect blocks */
2330         meta_blocks = 1;
2331         /* double indirect blocks */
2332         meta_blocks += 1 + (1LL << (bits-2));
2333         /* tripple indirect blocks */
2334         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2335
2336         upper_limit -= meta_blocks;
2337         upper_limit <<= bits;
2338
2339         res += 1LL << (bits-2);
2340         res += 1LL << (2*(bits-2));
2341         res += 1LL << (3*(bits-2));
2342         res <<= bits;
2343         if (res > upper_limit)
2344                 res = upper_limit;
2345
2346         if (res > MAX_LFS_FILESIZE)
2347                 res = MAX_LFS_FILESIZE;
2348
2349         return res;
2350 }
2351
2352 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2353                                    ext4_fsblk_t logical_sb_block, int nr)
2354 {
2355         struct ext4_sb_info *sbi = EXT4_SB(sb);
2356         ext4_group_t bg, first_meta_bg;
2357         int has_super = 0;
2358
2359         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2360
2361         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2362             nr < first_meta_bg)
2363                 return logical_sb_block + nr + 1;
2364         bg = sbi->s_desc_per_block * nr;
2365         if (ext4_bg_has_super(sb, bg))
2366                 has_super = 1;
2367
2368         /*
2369          * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2370          * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
2371          * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2372          * compensate.
2373          */
2374         if (sb->s_blocksize == 1024 && nr == 0 &&
2375             le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2376                 has_super++;
2377
2378         return (has_super + ext4_group_first_block_no(sb, bg));
2379 }
2380
2381 /**
2382  * ext4_get_stripe_size: Get the stripe size.
2383  * @sbi: In memory super block info
2384  *
2385  * If we have specified it via mount option, then
2386  * use the mount option value. If the value specified at mount time is
2387  * greater than the blocks per group use the super block value.
2388  * If the super block value is greater than blocks per group return 0.
2389  * Allocator needs it be less than blocks per group.
2390  *
2391  */
2392 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2393 {
2394         unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2395         unsigned long stripe_width =
2396                         le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2397         int ret;
2398
2399         if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2400                 ret = sbi->s_stripe;
2401         else if (stripe_width <= sbi->s_blocks_per_group)
2402                 ret = stripe_width;
2403         else if (stride <= sbi->s_blocks_per_group)
2404                 ret = stride;
2405         else
2406                 ret = 0;
2407
2408         /*
2409          * If the stripe width is 1, this makes no sense and
2410          * we set it to 0 to turn off stripe handling code.
2411          */
2412         if (ret <= 1)
2413                 ret = 0;
2414
2415         return ret;
2416 }
2417
2418 /* sysfs supprt */
2419
2420 struct ext4_attr {
2421         struct attribute attr;
2422         ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2423         ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2424                          const char *, size_t);
2425         union {
2426                 int offset;
2427                 int deprecated_val;
2428         } u;
2429 };
2430
2431 static int parse_strtoull(const char *buf,
2432                 unsigned long long max, unsigned long long *value)
2433 {
2434         int ret;
2435
2436         ret = kstrtoull(skip_spaces(buf), 0, value);
2437         if (!ret && *value > max)
2438                 ret = -EINVAL;
2439         return ret;
2440 }
2441
2442 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2443                                               struct ext4_sb_info *sbi,
2444                                               char *buf)
2445 {
2446         return snprintf(buf, PAGE_SIZE, "%llu\n",
2447                 (s64) EXT4_C2B(sbi,
2448                         percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2449 }
2450
2451 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2452                                          struct ext4_sb_info *sbi, char *buf)
2453 {
2454         struct super_block *sb = sbi->s_buddy_cache->i_sb;
2455
2456         if (!sb->s_bdev->bd_part)
2457                 return snprintf(buf, PAGE_SIZE, "0\n");
2458         return snprintf(buf, PAGE_SIZE, "%lu\n",
2459                         (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2460                          sbi->s_sectors_written_start) >> 1);
2461 }
2462
2463 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2464                                           struct ext4_sb_info *sbi, char *buf)
2465 {
2466         struct super_block *sb = sbi->s_buddy_cache->i_sb;
2467
2468         if (!sb->s_bdev->bd_part)
2469                 return snprintf(buf, PAGE_SIZE, "0\n");
2470         return snprintf(buf, PAGE_SIZE, "%llu\n",
2471                         (unsigned long long)(sbi->s_kbytes_written +
2472                         ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2473                           EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2474 }
2475
2476 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2477                                           struct ext4_sb_info *sbi,
2478                                           const char *buf, size_t count)
2479 {
2480         unsigned long t;
2481         int ret;
2482
2483         ret = kstrtoul(skip_spaces(buf), 0, &t);
2484         if (ret)
2485                 return ret;
2486
2487         if (t && (!is_power_of_2(t) || t > 0x40000000))
2488                 return -EINVAL;
2489
2490         sbi->s_inode_readahead_blks = t;
2491         return count;
2492 }
2493
2494 static ssize_t sbi_ui_show(struct ext4_attr *a,
2495                            struct ext4_sb_info *sbi, char *buf)
2496 {
2497         unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2498
2499         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2500 }
2501
2502 static ssize_t sbi_ui_store(struct ext4_attr *a,
2503                             struct ext4_sb_info *sbi,
2504                             const char *buf, size_t count)
2505 {
2506         unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2507         unsigned long t;
2508         int ret;
2509
2510         ret = kstrtoul(skip_spaces(buf), 0, &t);
2511         if (ret)
2512                 return ret;
2513         *ui = t;
2514         return count;
2515 }
2516
2517 static ssize_t es_ui_show(struct ext4_attr *a,
2518                            struct ext4_sb_info *sbi, char *buf)
2519 {
2520
2521         unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2522                            a->u.offset);
2523
2524         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2525 }
2526
2527 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2528                                   struct ext4_sb_info *sbi, char *buf)
2529 {
2530         return snprintf(buf, PAGE_SIZE, "%llu\n",
2531                 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2532 }
2533
2534 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2535                                    struct ext4_sb_info *sbi,
2536                                    const char *buf, size_t count)
2537 {
2538         unsigned long long val;
2539         int ret;
2540
2541         if (parse_strtoull(buf, -1ULL, &val))
2542                 return -EINVAL;
2543         ret = ext4_reserve_clusters(sbi, val);
2544
2545         return ret ? ret : count;
2546 }
2547
2548 static ssize_t trigger_test_error(struct ext4_attr *a,
2549                                   struct ext4_sb_info *sbi,
2550                                   const char *buf, size_t count)
2551 {
2552         int len = count;
2553
2554         if (!capable(CAP_SYS_ADMIN))
2555                 return -EPERM;
2556
2557         if (len && buf[len-1] == '\n')
2558                 len--;
2559
2560         if (len)
2561                 ext4_error(sbi->s_sb, "%.*s", len, buf);
2562         return count;
2563 }
2564
2565 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2566                                    struct ext4_sb_info *sbi, char *buf)
2567 {
2568         return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2569 }
2570
2571 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2572 static struct ext4_attr ext4_attr_##_name = {                   \
2573         .attr = {.name = __stringify(_name), .mode = _mode },   \
2574         .show   = _show,                                        \
2575         .store  = _store,                                       \
2576         .u = {                                                  \
2577                 .offset = offsetof(struct ext4_sb_info, _elname),\
2578         },                                                      \
2579 }
2580
2581 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname)           \
2582 static struct ext4_attr ext4_attr_##_name = {                           \
2583         .attr = {.name = __stringify(_name), .mode = _mode },           \
2584         .show   = _show,                                                \
2585         .store  = _store,                                               \
2586         .u = {                                                          \
2587                 .offset = offsetof(struct ext4_super_block, _elname),   \
2588         },                                                              \
2589 }
2590
2591 #define EXT4_ATTR(name, mode, show, store) \
2592 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2593
2594 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2595 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2596 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2597
2598 #define EXT4_RO_ATTR_ES_UI(name, elname)        \
2599         EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2600 #define EXT4_RW_ATTR_SBI_UI(name, elname)       \
2601         EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2602
2603 #define ATTR_LIST(name) &ext4_attr_##name.attr
2604 #define EXT4_DEPRECATED_ATTR(_name, _val)       \
2605 static struct ext4_attr ext4_attr_##_name = {                   \
2606         .attr = {.name = __stringify(_name), .mode = 0444 },    \
2607         .show   = sbi_deprecated_show,                          \
2608         .u = {                                                  \
2609                 .deprecated_val = _val,                         \
2610         },                                                      \
2611 }
2612
2613 EXT4_RO_ATTR(delayed_allocation_blocks);
2614 EXT4_RO_ATTR(session_write_kbytes);
2615 EXT4_RO_ATTR(lifetime_write_kbytes);
2616 EXT4_RW_ATTR(reserved_clusters);
2617 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2618                  inode_readahead_blks_store, s_inode_readahead_blks);
2619 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2620 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2621 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2622 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2623 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2624 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2625 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2626 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2627 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2628 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2629 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2630 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2631 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2632 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2633 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2634 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2635 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2636 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2637 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2638
2639 static struct attribute *ext4_attrs[] = {
2640         ATTR_LIST(delayed_allocation_blocks),
2641         ATTR_LIST(session_write_kbytes),
2642         ATTR_LIST(lifetime_write_kbytes),
2643         ATTR_LIST(reserved_clusters),
2644         ATTR_LIST(inode_readahead_blks),
2645         ATTR_LIST(inode_goal),
2646         ATTR_LIST(mb_stats),
2647         ATTR_LIST(mb_max_to_scan),
2648         ATTR_LIST(mb_min_to_scan),
2649         ATTR_LIST(mb_order2_req),
2650         ATTR_LIST(mb_stream_req),
2651         ATTR_LIST(mb_group_prealloc),
2652         ATTR_LIST(max_writeback_mb_bump),
2653         ATTR_LIST(extent_max_zeroout_kb),
2654         ATTR_LIST(trigger_fs_error),
2655         ATTR_LIST(err_ratelimit_interval_ms),
2656         ATTR_LIST(err_ratelimit_burst),
2657         ATTR_LIST(warning_ratelimit_interval_ms),
2658         ATTR_LIST(warning_ratelimit_burst),
2659         ATTR_LIST(msg_ratelimit_interval_ms),
2660         ATTR_LIST(msg_ratelimit_burst),
2661         ATTR_LIST(errors_count),
2662         ATTR_LIST(first_error_time),
2663         ATTR_LIST(last_error_time),
2664         NULL,
2665 };
2666
2667 /* Features this copy of ext4 supports */
2668 EXT4_INFO_ATTR(lazy_itable_init);
2669 EXT4_INFO_ATTR(batched_discard);
2670 EXT4_INFO_ATTR(meta_bg_resize);
2671
2672 static struct attribute *ext4_feat_attrs[] = {
2673         ATTR_LIST(lazy_itable_init),
2674         ATTR_LIST(batched_discard),
2675         ATTR_LIST(meta_bg_resize),
2676         NULL,
2677 };
2678
2679 static ssize_t ext4_attr_show(struct kobject *kobj,
2680                               struct attribute *attr, char *buf)
2681 {
2682         struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2683                                                 s_kobj);
2684         struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2685
2686         return a->show ? a->show(a, sbi, buf) : 0;
2687 }
2688
2689 static ssize_t ext4_attr_store(struct kobject *kobj,
2690                                struct attribute *attr,
2691                                const char *buf, size_t len)
2692 {
2693         struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2694                                                 s_kobj);
2695         struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2696
2697         return a->store ? a->store(a, sbi, buf, len) : 0;
2698 }
2699
2700 static void ext4_sb_release(struct kobject *kobj)
2701 {
2702         struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2703                                                 s_kobj);
2704         complete(&sbi->s_kobj_unregister);
2705 }
2706
2707 static const struct sysfs_ops ext4_attr_ops = {
2708         .show   = ext4_attr_show,
2709         .store  = ext4_attr_store,
2710 };
2711
2712 static struct kobj_type ext4_ktype = {
2713         .default_attrs  = ext4_attrs,
2714         .sysfs_ops      = &ext4_attr_ops,
2715         .release        = ext4_sb_release,
2716 };
2717
2718 static void ext4_feat_release(struct kobject *kobj)
2719 {
2720         complete(&ext4_feat->f_kobj_unregister);
2721 }
2722
2723 static ssize_t ext4_feat_show(struct kobject *kobj,
2724                               struct attribute *attr, char *buf)
2725 {
2726         return snprintf(buf, PAGE_SIZE, "supported\n");
2727 }
2728
2729 /*
2730  * We can not use ext4_attr_show/store because it relies on the kobject
2731  * being embedded in the ext4_sb_info structure which is definitely not
2732  * true in this case.
2733  */
2734 static const struct sysfs_ops ext4_feat_ops = {
2735         .show   = ext4_feat_show,
2736         .store  = NULL,
2737 };
2738
2739 static struct kobj_type ext4_feat_ktype = {
2740         .default_attrs  = ext4_feat_attrs,
2741         .sysfs_ops      = &ext4_feat_ops,
2742         .release        = ext4_feat_release,
2743 };
2744
2745 /*
2746  * Check whether this filesystem can be mounted based on
2747  * the features present and the RDONLY/RDWR mount requested.
2748  * Returns 1 if this filesystem can be mounted as requested,
2749  * 0 if it cannot be.
2750  */
2751 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2752 {
2753         if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2754                 ext4_msg(sb, KERN_ERR,
2755                         "Couldn't mount because of "
2756                         "unsupported optional features (%x)",
2757                         (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2758                         ~EXT4_FEATURE_INCOMPAT_SUPP));
2759                 return 0;
2760         }
2761
2762         if (readonly)
2763                 return 1;
2764
2765         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2766                 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2767                 sb->s_flags |= MS_RDONLY;
2768                 return 1;
2769         }
2770
2771         /* Check that feature set is OK for a read-write mount */
2772         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2773                 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2774                          "unsupported optional features (%x)",
2775                          (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2776                                 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2777                 return 0;
2778         }
2779         /*
2780          * Large file size enabled file system can only be mounted
2781          * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2782          */
2783         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2784                 if (sizeof(blkcnt_t) < sizeof(u64)) {
2785                         ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2786                                  "cannot be mounted RDWR without "
2787                                  "CONFIG_LBDAF");
2788                         return 0;
2789                 }
2790         }
2791         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2792             !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2793                 ext4_msg(sb, KERN_ERR,
2794                          "Can't support bigalloc feature without "
2795                          "extents feature\n");
2796                 return 0;
2797         }
2798
2799 #ifndef CONFIG_QUOTA
2800         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2801             !readonly) {
2802                 ext4_msg(sb, KERN_ERR,
2803                          "Filesystem with quota feature cannot be mounted RDWR "
2804                          "without CONFIG_QUOTA");
2805                 return 0;
2806         }
2807 #endif  /* CONFIG_QUOTA */
2808         return 1;
2809 }
2810
2811 /*
2812  * This function is called once a day if we have errors logged
2813  * on the file system
2814  */
2815 static void print_daily_error_info(unsigned long arg)
2816 {
2817         struct super_block *sb = (struct super_block *) arg;
2818         struct ext4_sb_info *sbi;
2819         struct ext4_super_block *es;
2820
2821         sbi = EXT4_SB(sb);
2822         es = sbi->s_es;
2823
2824         if (es->s_error_count)
2825                 /* fsck newer than v1.41.13 is needed to clean this condition. */
2826                 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2827                          le32_to_cpu(es->s_error_count));
2828         if (es->s_first_error_time) {
2829                 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2830                        sb->s_id, le32_to_cpu(es->s_first_error_time),
2831                        (int) sizeof(es->s_first_error_func),
2832                        es->s_first_error_func,
2833                        le32_to_cpu(es->s_first_error_line));
2834                 if (es->s_first_error_ino)
2835                         printk(": inode %u",
2836                                le32_to_cpu(es->s_first_error_ino));
2837                 if (es->s_first_error_block)
2838                         printk(": block %llu", (unsigned long long)
2839                                le64_to_cpu(es->s_first_error_block));
2840                 printk("\n");
2841         }
2842         if (es->s_last_error_time) {
2843                 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2844                        sb->s_id, le32_to_cpu(es->s_last_error_time),
2845                        (int) sizeof(es->s_last_error_func),
2846                        es->s_last_error_func,
2847                        le32_to_cpu(es->s_last_error_line));
2848                 if (es->s_last_error_ino)
2849                         printk(": inode %u",
2850                                le32_to_cpu(es->s_last_error_ino));
2851                 if (es->s_last_error_block)
2852                         printk(": block %llu", (unsigned long long)
2853                                le64_to_cpu(es->s_last_error_block));
2854                 printk("\n");
2855         }
2856         mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2857 }
2858
2859 /* Find next suitable group and run ext4_init_inode_table */
2860 static int ext4_run_li_request(struct ext4_li_request *elr)
2861 {
2862         struct ext4_group_desc *gdp = NULL;
2863         ext4_group_t group, ngroups;
2864         struct super_block *sb;
2865         unsigned long timeout = 0;
2866         int ret = 0;
2867
2868         sb = elr->lr_super;
2869         ngroups = EXT4_SB(sb)->s_groups_count;
2870
2871         sb_start_write(sb);
2872         for (group = elr->lr_next_group; group < ngroups; group++) {
2873                 gdp = ext4_get_group_desc(sb, group, NULL);
2874                 if (!gdp) {
2875                         ret = 1;
2876                         break;
2877                 }
2878
2879                 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2880                         break;
2881         }
2882
2883         if (group >= ngroups)
2884                 ret = 1;
2885
2886         if (!ret) {
2887                 timeout = jiffies;
2888                 ret = ext4_init_inode_table(sb, group,
2889                                             elr->lr_timeout ? 0 : 1);
2890                 if (elr->lr_timeout == 0) {
2891                         timeout = (jiffies - timeout) *
2892                                   elr->lr_sbi->s_li_wait_mult;
2893                         elr->lr_timeout = timeout;
2894                 }
2895                 elr->lr_next_sched = jiffies + elr->lr_timeout;
2896                 elr->lr_next_group = group + 1;
2897         }
2898         sb_end_write(sb);
2899
2900         return ret;
2901 }
2902
2903 /*
2904  * Remove lr_request from the list_request and free the
2905  * request structure. Should be called with li_list_mtx held
2906  */
2907 static void ext4_remove_li_request(struct ext4_li_request *elr)
2908 {
2909         struct ext4_sb_info *sbi;
2910
2911         if (!elr)
2912                 return;
2913
2914         sbi = elr->lr_sbi;
2915
2916         list_del(&elr->lr_request);
2917         sbi->s_li_request = NULL;
2918         kfree(elr);
2919 }
2920
2921 static void ext4_unregister_li_request(struct super_block *sb)
2922 {
2923         mutex_lock(&ext4_li_mtx);
2924         if (!ext4_li_info) {
2925                 mutex_unlock(&ext4_li_mtx);
2926                 return;
2927         }
2928
2929         mutex_lock(&ext4_li_info->li_list_mtx);
2930         ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2931         mutex_unlock(&ext4_li_info->li_list_mtx);
2932         mutex_unlock(&ext4_li_mtx);
2933 }
2934
2935 static struct task_struct *ext4_lazyinit_task;
2936
2937 /*
2938  * This is the function where ext4lazyinit thread lives. It walks
2939  * through the request list searching for next scheduled filesystem.
2940  * When such a fs is found, run the lazy initialization request
2941  * (ext4_rn_li_request) and keep track of the time spend in this
2942  * function. Based on that time we compute next schedule time of
2943  * the request. When walking through the list is complete, compute
2944  * next waking time and put itself into sleep.
2945  */
2946 static int ext4_lazyinit_thread(void *arg)
2947 {
2948         struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2949         struct list_head *pos, *n;
2950         struct ext4_li_request *elr;
2951         unsigned long next_wakeup, cur;
2952
2953         BUG_ON(NULL == eli);
2954
2955 cont_thread:
2956         while (true) {
2957                 next_wakeup = MAX_JIFFY_OFFSET;
2958
2959                 mutex_lock(&eli->li_list_mtx);
2960                 if (list_empty(&eli->li_request_list)) {
2961                         mutex_unlock(&eli->li_list_mtx);
2962                         goto exit_thread;
2963                 }
2964
2965                 list_for_each_safe(pos, n, &eli->li_request_list) {
2966                         elr = list_entry(pos, struct ext4_li_request,
2967                                          lr_request);
2968
2969                         if (time_after_eq(jiffies, elr->lr_next_sched)) {
2970                                 if (ext4_run_li_request(elr) != 0) {
2971                                         /* error, remove the lazy_init job */
2972                                         ext4_remove_li_request(elr);
2973                                         continue;
2974                                 }
2975                         }
2976
2977                         if (time_before(elr->lr_next_sched, next_wakeup))
2978                                 next_wakeup = elr->lr_next_sched;
2979                 }
2980                 mutex_unlock(&eli->li_list_mtx);
2981
2982                 try_to_freeze();
2983
2984                 cur = jiffies;
2985                 if ((time_after_eq(cur, next_wakeup)) ||
2986                     (MAX_JIFFY_OFFSET == next_wakeup)) {
2987                         cond_resched();
2988                         continue;
2989                 }
2990
2991                 schedule_timeout_interruptible(next_wakeup - cur);
2992
2993                 if (kthread_should_stop()) {
2994                         ext4_clear_request_list();
2995                         goto exit_thread;
2996                 }
2997         }
2998
2999 exit_thread:
3000         /*
3001          * It looks like the request list is empty, but we need
3002          * to check it under the li_list_mtx lock, to prevent any
3003          * additions into it, and of course we should lock ext4_li_mtx
3004          * to atomically free the list and ext4_li_info, because at
3005          * this point another ext4 filesystem could be registering
3006          * new one.
3007          */
3008         mutex_lock(&ext4_li_mtx);
3009         mutex_lock(&eli->li_list_mtx);
3010         if (!list_empty(&eli->li_request_list)) {
3011                 mutex_unlock(&eli->li_list_mtx);
3012                 mutex_unlock(&ext4_li_mtx);
3013                 goto cont_thread;
3014         }
3015         mutex_unlock(&eli->li_list_mtx);
3016         kfree(ext4_li_info);
3017         ext4_li_info = NULL;
3018         mutex_unlock(&ext4_li_mtx);
3019
3020         return 0;
3021 }
3022
3023 static void ext4_clear_request_list(void)
3024 {
3025         struct list_head *pos, *n;
3026         struct ext4_li_request *elr;
3027
3028         mutex_lock(&ext4_li_info->li_list_mtx);
3029         list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3030                 elr = list_entry(pos, struct ext4_li_request,
3031                                  lr_request);
3032                 ext4_remove_li_request(elr);
3033         }
3034         mutex_unlock(&ext4_li_info->li_list_mtx);
3035 }
3036
3037 static int ext4_run_lazyinit_thread(void)
3038 {
3039         ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3040                                          ext4_li_info, "ext4lazyinit");
3041         if (IS_ERR(ext4_lazyinit_task)) {
3042                 int err = PTR_ERR(ext4_lazyinit_task);
3043                 ext4_clear_request_list();
3044                 kfree(ext4_li_info);
3045                 ext4_li_info = NULL;
3046                 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3047                                  "initialization thread\n",
3048                                  err);
3049                 return err;
3050         }
3051         ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3052         return 0;
3053 }
3054
3055 /*
3056  * Check whether it make sense to run itable init. thread or not.
3057  * If there is at least one uninitialized inode table, return
3058  * corresponding group number, else the loop goes through all
3059  * groups and return total number of groups.
3060  */
3061 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3062 {
3063         ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3064         struct ext4_group_desc *gdp = NULL;
3065
3066         for (group = 0; group < ngroups; group++) {
3067                 gdp = ext4_get_group_desc(sb, group, NULL);
3068                 if (!gdp)
3069                         continue;
3070
3071                 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3072                         break;
3073         }
3074
3075         return group;
3076 }
3077
3078 static int ext4_li_info_new(void)
3079 {
3080         struct ext4_lazy_init *eli = NULL;
3081
3082         eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3083         if (!eli)
3084                 return -ENOMEM;
3085
3086         INIT_LIST_HEAD(&eli->li_request_list);
3087         mutex_init(&eli->li_list_mtx);
3088
3089         eli->li_state |= EXT4_LAZYINIT_QUIT;
3090
3091         ext4_li_info = eli;
3092
3093         return 0;
3094 }
3095
3096 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3097                                             ext4_group_t start)
3098 {
3099         struct ext4_sb_info *sbi = EXT4_SB(sb);
3100         struct ext4_li_request *elr;
3101
3102         elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3103         if (!elr)
3104                 return NULL;
3105
3106         elr->lr_super = sb;
3107         elr->lr_sbi = sbi;
3108         elr->lr_next_group = start;
3109
3110         /*
3111          * Randomize first schedule time of the request to
3112          * spread the inode table initialization requests
3113          * better.
3114          */
3115         elr->lr_next_sched = jiffies + (prandom_u32() %
3116                                 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3117         return elr;
3118 }
3119
3120 int ext4_register_li_request(struct super_block *sb,
3121                              ext4_group_t first_not_zeroed)
3122 {
3123         struct ext4_sb_info *sbi = EXT4_SB(sb);
3124         struct ext4_li_request *elr = NULL;
3125         ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3126         int ret = 0;
3127
3128         mutex_lock(&ext4_li_mtx);
3129         if (sbi->s_li_request != NULL) {
3130                 /*
3131                  * Reset timeout so it can be computed again, because
3132                  * s_li_wait_mult might have changed.
3133                  */
3134                 sbi->s_li_request->lr_timeout = 0;
3135                 goto out;
3136         }
3137
3138         if (first_not_zeroed == ngroups ||
3139             (sb->s_flags & MS_RDONLY) ||
3140             !test_opt(sb, INIT_INODE_TABLE))
3141                 goto out;
3142
3143         elr = ext4_li_request_new(sb, first_not_zeroed);
3144         if (!elr) {
3145                 ret = -ENOMEM;
3146                 goto out;
3147         }
3148
3149         if (NULL == ext4_li_info) {
3150                 ret = ext4_li_info_new();
3151                 if (ret)
3152                         goto out;
3153         }
3154
3155         mutex_lock(&ext4_li_info->li_list_mtx);
3156         list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3157         mutex_unlock(&ext4_li_info->li_list_mtx);
3158
3159         sbi->s_li_request = elr;
3160         /*
3161          * set elr to NULL here since it has been inserted to
3162          * the request_list and the removal and free of it is
3163          * handled by ext4_clear_request_list from now on.
3164          */
3165         elr = NULL;
3166
3167         if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3168                 ret = ext4_run_lazyinit_thread();
3169                 if (ret)
3170                         goto out;
3171         }
3172 out:
3173         mutex_unlock(&ext4_li_mtx);
3174         if (ret)
3175                 kfree(elr);
3176         return ret;
3177 }
3178
3179 /*
3180  * We do not need to lock anything since this is called on
3181  * module unload.
3182  */
3183 static void ext4_destroy_lazyinit_thread(void)
3184 {
3185         /*
3186          * If thread exited earlier
3187          * there's nothing to be done.
3188          */
3189         if (!ext4_li_info || !ext4_lazyinit_task)
3190                 return;
3191
3192         kthread_stop(ext4_lazyinit_task);
3193 }
3194
3195 static int set_journal_csum_feature_set(struct super_block *sb)
3196 {
3197         int ret = 1;
3198         int compat, incompat;
3199         struct ext4_sb_info *sbi = EXT4_SB(sb);
3200
3201         if (ext4_has_metadata_csum(sb)) {
3202                 /* journal checksum v3 */
3203                 compat = 0;
3204                 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3205         } else {
3206                 /* journal checksum v1 */
3207                 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3208                 incompat = 0;
3209         }
3210
3211         jbd2_journal_clear_features(sbi->s_journal,
3212                         JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3213                         JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3214                         JBD2_FEATURE_INCOMPAT_CSUM_V2);
3215         if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3216                 ret = jbd2_journal_set_features(sbi->s_journal,
3217                                 compat, 0,
3218                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3219                                 incompat);
3220         } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3221                 ret = jbd2_journal_set_features(sbi->s_journal,
3222                                 compat, 0,
3223                                 incompat);
3224                 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3225                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3226         } else {
3227                 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3228                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3229         }
3230
3231         return ret;
3232 }
3233
3234 /*
3235  * Note: calculating the overhead so we can be compatible with
3236  * historical BSD practice is quite difficult in the face of
3237  * clusters/bigalloc.  This is because multiple metadata blocks from
3238  * different block group can end up in the same allocation cluster.
3239  * Calculating the exact overhead in the face of clustered allocation
3240  * requires either O(all block bitmaps) in memory or O(number of block
3241  * groups**2) in time.  We will still calculate the superblock for
3242  * older file systems --- and if we come across with a bigalloc file
3243  * system with zero in s_overhead_clusters the estimate will be close to
3244  * correct especially for very large cluster sizes --- but for newer
3245  * file systems, it's better to calculate this figure once at mkfs
3246  * time, and store it in the superblock.  If the superblock value is
3247  * present (even for non-bigalloc file systems), we will use it.
3248  */
3249 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3250                           char *buf)
3251 {
3252         struct ext4_sb_info     *sbi = EXT4_SB(sb);
3253         struct ext4_group_desc  *gdp;
3254         ext4_fsblk_t            first_block, last_block, b;
3255         ext4_group_t            i, ngroups = ext4_get_groups_count(sb);
3256         int                     s, j, count = 0;
3257
3258         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3259                 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3260                         sbi->s_itb_per_group + 2);
3261
3262         first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3263                 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3264         last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3265         for (i = 0; i < ngroups; i++) {
3266                 gdp = ext4_get_group_desc(sb, i, NULL);
3267                 b = ext4_block_bitmap(sb, gdp);
3268                 if (b >= first_block && b <= last_block) {
3269                         ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3270                         count++;
3271                 }
3272                 b = ext4_inode_bitmap(sb, gdp);
3273                 if (b >= first_block && b <= last_block) {
3274                         ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3275                         count++;
3276                 }
3277                 b = ext4_inode_table(sb, gdp);
3278                 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3279                         for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3280                                 int c = EXT4_B2C(sbi, b - first_block);
3281                                 ext4_set_bit(c, buf);
3282                                 count++;
3283                         }
3284                 if (i != grp)
3285                         continue;
3286                 s = 0;
3287                 if (ext4_bg_has_super(sb, grp)) {
3288                         ext4_set_bit(s++, buf);
3289                         count++;
3290                 }
3291                 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3292                         ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3293                         count++;
3294                 }
3295         }
3296         if (!count)
3297                 return 0;
3298         return EXT4_CLUSTERS_PER_GROUP(sb) -
3299                 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3300 }
3301
3302 /*
3303  * Compute the overhead and stash it in sbi->s_overhead
3304  */
3305 int ext4_calculate_overhead(struct super_block *sb)
3306 {
3307         struct ext4_sb_info *sbi = EXT4_SB(sb);
3308         struct ext4_super_block *es = sbi->s_es;
3309         ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3310         ext4_fsblk_t overhead = 0;
3311         char *buf = (char *) get_zeroed_page(GFP_NOFS);
3312
3313         if (!buf)
3314                 return -ENOMEM;
3315
3316         /*
3317          * Compute the overhead (FS structures).  This is constant
3318          * for a given filesystem unless the number of block groups
3319          * changes so we cache the previous value until it does.
3320          */
3321
3322         /*
3323          * All of the blocks before first_data_block are overhead
3324          */
3325         overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3326
3327         /*
3328          * Add the overhead found in each block group
3329          */
3330         for (i = 0; i < ngroups; i++) {
3331                 int blks;
3332
3333                 blks = count_overhead(sb, i, buf);
3334                 overhead += blks;
3335                 if (blks)
3336                         memset(buf, 0, PAGE_SIZE);
3337                 cond_resched();
3338         }
3339         /* Add the internal journal blocks as well */
3340         if (sbi->s_journal && !sbi->journal_bdev)
3341                 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3342
3343         sbi->s_overhead = overhead;
3344         smp_wmb();
3345         free_page((unsigned long) buf);
3346         return 0;
3347 }
3348
3349
3350 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3351 {
3352         ext4_fsblk_t resv_clusters;
3353
3354         /*
3355          * There's no need to reserve anything when we aren't using extents.
3356          * The space estimates are exact, there are no unwritten extents,
3357          * hole punching doesn't need new metadata... This is needed especially
3358          * to keep ext2/3 backward compatibility.
3359          */
3360         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3361                 return 0;
3362         /*
3363          * By default we reserve 2% or 4096 clusters, whichever is smaller.
3364          * This should cover the situations where we can not afford to run
3365          * out of space like for example punch hole, or converting
3366          * unwritten extents in delalloc path. In most cases such
3367          * allocation would require 1, or 2 blocks, higher numbers are
3368          * very rare.
3369          */
3370         resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3371                         EXT4_SB(sb)->s_cluster_bits;
3372
3373         do_div(resv_clusters, 50);
3374         resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3375
3376         return resv_clusters;
3377 }
3378
3379
3380 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3381 {
3382         ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3383                                 sbi->s_cluster_bits;
3384
3385         if (count >= clusters)
3386                 return -EINVAL;
3387
3388         atomic64_set(&sbi->s_resv_clusters, count);
3389         return 0;
3390 }
3391
3392 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3393 {
3394         char *orig_data = kstrdup(data, GFP_KERNEL);
3395         struct buffer_head *bh;
3396         struct ext4_super_block *es = NULL;
3397         struct ext4_sb_info *sbi;
3398         ext4_fsblk_t block;
3399         ext4_fsblk_t sb_block = get_sb_block(&data);
3400         ext4_fsblk_t logical_sb_block;
3401         unsigned long offset = 0;
3402         unsigned long journal_devnum = 0;
3403         unsigned long def_mount_opts;
3404         struct inode *root;
3405         char *cp;
3406         const char *descr;
3407         int ret = -ENOMEM;
3408         int blocksize, clustersize;
3409         unsigned int db_count;
3410         unsigned int i;
3411         int needs_recovery, has_huge_files, has_bigalloc;
3412         __u64 blocks_count;
3413         int err = 0;
3414         unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3415         ext4_group_t first_not_zeroed;
3416
3417         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3418         if (!sbi)
3419                 goto out_free_orig;
3420
3421         sbi->s_blockgroup_lock =
3422                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3423         if (!sbi->s_blockgroup_lock) {
3424                 kfree(sbi);
3425                 goto out_free_orig;
3426         }
3427         sb->s_fs_info = sbi;
3428         sbi->s_sb = sb;
3429         sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3430         sbi->s_sb_block = sb_block;
3431         if (sb->s_bdev->bd_part)
3432                 sbi->s_sectors_written_start =
3433                         part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3434
3435         /* Cleanup superblock name */
3436         for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3437                 *cp = '!';
3438
3439         /* -EINVAL is default */
3440         ret = -EINVAL;
3441         blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3442         if (!blocksize) {
3443                 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3444                 goto out_fail;
3445         }
3446
3447         /*
3448          * The ext4 superblock will not be buffer aligned for other than 1kB
3449          * block sizes.  We need to calculate the offset from buffer start.
3450          */
3451         if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3452                 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3453                 offset = do_div(logical_sb_block, blocksize);
3454         } else {
3455                 logical_sb_block = sb_block;
3456         }
3457
3458         if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3459                 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3460                 goto out_fail;
3461         }
3462         /*
3463          * Note: s_es must be initialized as soon as possible because
3464          *       some ext4 macro-instructions depend on its value
3465          */
3466         es = (struct ext4_super_block *) (bh->b_data + offset);
3467         sbi->s_es = es;
3468         sb->s_magic = le16_to_cpu(es->s_magic);
3469         if (sb->s_magic != EXT4_SUPER_MAGIC)
3470                 goto cantfind_ext4;
3471         sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3472
3473         /* Warn if metadata_csum and gdt_csum are both set. */
3474         if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3475                                        EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3476             EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3477                 ext4_warning(sb, "metadata_csum and uninit_bg are "
3478                              "redundant flags; please run fsck.");
3479
3480         /* Check for a known checksum algorithm */
3481         if (!ext4_verify_csum_type(sb, es)) {
3482                 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3483                          "unknown checksum algorithm.");
3484                 silent = 1;
3485                 goto cantfind_ext4;
3486         }
3487
3488         /* Load the checksum driver */
3489         if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3490                                        EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3491                 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3492                 if (IS_ERR(sbi->s_chksum_driver)) {
3493                         ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3494                         ret = PTR_ERR(sbi->s_chksum_driver);
3495                         sbi->s_chksum_driver = NULL;
3496                         goto failed_mount;
3497                 }
3498         }
3499
3500         /* Check superblock checksum */
3501         if (!ext4_superblock_csum_verify(sb, es)) {
3502                 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3503                          "invalid superblock checksum.  Run e2fsck?");
3504                 silent = 1;
3505                 goto cantfind_ext4;
3506         }
3507
3508         /* Precompute checksum seed for all metadata */
3509         if (ext4_has_metadata_csum(sb))
3510                 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3511                                                sizeof(es->s_uuid));
3512
3513         /* Set defaults before we parse the mount options */
3514         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3515         set_opt(sb, INIT_INODE_TABLE);
3516         if (def_mount_opts & EXT4_DEFM_DEBUG)
3517                 set_opt(sb, DEBUG);
3518         if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3519                 set_opt(sb, GRPID);
3520         if (def_mount_opts & EXT4_DEFM_UID16)
3521                 set_opt(sb, NO_UID32);
3522         /* xattr user namespace & acls are now defaulted on */
3523         set_opt(sb, XATTR_USER);
3524 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3525         set_opt(sb, POSIX_ACL);
3526 #endif
3527         /* don't forget to enable journal_csum when metadata_csum is enabled. */
3528         if (ext4_has_metadata_csum(sb))
3529                 set_opt(sb, JOURNAL_CHECKSUM);
3530
3531         if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3532                 set_opt(sb, JOURNAL_DATA);
3533         else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3534                 set_opt(sb, ORDERED_DATA);
3535         else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3536                 set_opt(sb, WRITEBACK_DATA);
3537
3538         if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3539                 set_opt(sb, ERRORS_PANIC);
3540         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3541                 set_opt(sb, ERRORS_CONT);
3542         else
3543                 set_opt(sb, ERRORS_RO);
3544         /* block_validity enabled by default; disable with noblock_validity */
3545         set_opt(sb, BLOCK_VALIDITY);
3546         if (def_mount_opts & EXT4_DEFM_DISCARD)
3547                 set_opt(sb, DISCARD);
3548
3549         sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3550         sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3551         sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3552         sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3553         sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3554
3555         if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3556                 set_opt(sb, BARRIER);
3557
3558         /*
3559          * enable delayed allocation by default
3560          * Use -o nodelalloc to turn it off
3561          */
3562         if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3563             ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3564                 set_opt(sb, DELALLOC);
3565
3566         /*
3567          * set default s_li_wait_mult for lazyinit, for the case there is
3568          * no mount option specified.
3569          */
3570         sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3571
3572         if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3573                            &journal_devnum, &journal_ioprio, 0)) {
3574                 ext4_msg(sb, KERN_WARNING,
3575                          "failed to parse options in superblock: %s",
3576                          sbi->s_es->s_mount_opts);
3577         }
3578         sbi->s_def_mount_opt = sbi->s_mount_opt;
3579         if (!parse_options((char *) data, sb, &journal_devnum,
3580                            &journal_ioprio, 0))
3581                 goto failed_mount;
3582
3583         if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3584                 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3585                             "with data=journal disables delayed "
3586                             "allocation and O_DIRECT support!\n");
3587                 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3588                         ext4_msg(sb, KERN_ERR, "can't mount with "
3589                                  "both data=journal and delalloc");
3590                         goto failed_mount;
3591                 }
3592                 if (test_opt(sb, DIOREAD_NOLOCK)) {
3593                         ext4_msg(sb, KERN_ERR, "can't mount with "
3594                                  "both data=journal and dioread_nolock");
3595                         goto failed_mount;
3596                 }
3597                 if (test_opt(sb, DAX)) {
3598                         ext4_msg(sb, KERN_ERR, "can't mount with "
3599                                  "both data=journal and dax");
3600                         goto failed_mount;
3601                 }
3602                 if (test_opt(sb, DELALLOC))
3603                         clear_opt(sb, DELALLOC);
3604         }
3605
3606         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3607                 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3608
3609         if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3610             (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3611              EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3612              EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3613                 ext4_msg(sb, KERN_WARNING,
3614                        "feature flags set on rev 0 fs, "
3615                        "running e2fsck is recommended");
3616
3617         if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3618                 set_opt2(sb, HURD_COMPAT);
3619                 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3620                                               EXT4_FEATURE_INCOMPAT_64BIT)) {
3621                         ext4_msg(sb, KERN_ERR,
3622                                  "The Hurd can't support 64-bit file systems");
3623                         goto failed_mount;
3624                 }
3625         }
3626
3627         if (IS_EXT2_SB(sb)) {
3628                 if (ext2_feature_set_ok(sb))
3629                         ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3630                                  "using the ext4 subsystem");
3631                 else {
3632                         ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3633                                  "to feature incompatibilities");
3634                         goto failed_mount;
3635                 }
3636         }
3637
3638         if (IS_EXT3_SB(sb)) {
3639                 if (ext3_feature_set_ok(sb))
3640                         ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3641                                  "using the ext4 subsystem");
3642                 else {
3643                         ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3644                                  "to feature incompatibilities");
3645                         goto failed_mount;
3646                 }
3647         }
3648
3649         /*
3650          * Check feature flags regardless of the revision level, since we
3651          * previously didn't change the revision level when setting the flags,
3652          * so there is a chance incompat flags are set on a rev 0 filesystem.
3653          */
3654         if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3655                 goto failed_mount;
3656
3657         blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3658         if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3659             blocksize > EXT4_MAX_BLOCK_SIZE) {
3660                 ext4_msg(sb, KERN_ERR,
3661                        "Unsupported filesystem blocksize %d", blocksize);
3662                 goto failed_mount;
3663         }
3664
3665         if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3666                 if (blocksize != PAGE_SIZE) {
3667                         ext4_msg(sb, KERN_ERR,
3668                                         "error: unsupported blocksize for dax");
3669                         goto failed_mount;
3670                 }
3671                 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3672                         ext4_msg(sb, KERN_ERR,
3673                                         "error: device does not support dax");
3674                         goto failed_mount;
3675                 }
3676         }
3677
3678         if (sb->s_blocksize != blocksize) {
3679                 /* Validate the filesystem blocksize */
3680                 if (!sb_set_blocksize(sb, blocksize)) {
3681                         ext4_msg(sb, KERN_ERR, "bad block size %d",
3682                                         blocksize);
3683                         goto failed_mount;
3684                 }
3685
3686                 brelse(bh);
3687                 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3688                 offset = do_div(logical_sb_block, blocksize);
3689                 bh = sb_bread_unmovable(sb, logical_sb_block);
3690                 if (!bh) {
3691                         ext4_msg(sb, KERN_ERR,
3692                                "Can't read superblock on 2nd try");
3693                         goto failed_mount;
3694                 }
3695                 es = (struct ext4_super_block *)(bh->b_data + offset);
3696                 sbi->s_es = es;
3697                 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3698                         ext4_msg(sb, KERN_ERR,
3699                                "Magic mismatch, very weird!");
3700                         goto failed_mount;
3701                 }
3702         }
3703
3704         has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3705                                 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3706         sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3707                                                       has_huge_files);
3708         sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3709
3710         if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3711                 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3712                 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3713         } else {
3714                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3715                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3716                 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3717                     (!is_power_of_2(sbi->s_inode_size)) ||
3718                     (sbi->s_inode_size > blocksize)) {
3719                         ext4_msg(sb, KERN_ERR,
3720                                "unsupported inode size: %d",
3721                                sbi->s_inode_size);
3722                         goto failed_mount;
3723                 }
3724                 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3725                         sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3726         }
3727
3728         sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3729         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3730                 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3731                     sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3732                     !is_power_of_2(sbi->s_desc_size)) {
3733                         ext4_msg(sb, KERN_ERR,
3734                                "unsupported descriptor size %lu",
3735                                sbi->s_desc_size);
3736                         goto failed_mount;
3737                 }
3738         } else
3739                 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3740
3741         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3742         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3743         if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3744                 goto cantfind_ext4;
3745
3746         sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3747         if (sbi->s_inodes_per_block == 0)
3748                 goto cantfind_ext4;
3749         sbi->s_itb_per_group = sbi->s_inodes_per_group /
3750                                         sbi->s_inodes_per_block;
3751         sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3752         sbi->s_sbh = bh;
3753         sbi->s_mount_state = le16_to_cpu(es->s_state);
3754         sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3755         sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3756
3757         for (i = 0; i < 4; i++)
3758                 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3759         sbi->s_def_hash_version = es->s_def_hash_version;
3760         if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3761                 i = le32_to_cpu(es->s_flags);
3762                 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3763                         sbi->s_hash_unsigned = 3;
3764                 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3765 #ifdef __CHAR_UNSIGNED__
3766                         if (!(sb->s_flags & MS_RDONLY))
3767                                 es->s_flags |=
3768                                         cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3769                         sbi->s_hash_unsigned = 3;
3770 #else
3771                         if (!(sb->s_flags & MS_RDONLY))
3772                                 es->s_flags |=
3773                                         cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3774 #endif
3775                 }
3776         }
3777
3778         /* Handle clustersize */
3779         clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3780         has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3781                                 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3782         if (has_bigalloc) {
3783                 if (clustersize < blocksize) {
3784                         ext4_msg(sb, KERN_ERR,
3785                                  "cluster size (%d) smaller than "
3786                                  "block size (%d)", clustersize, blocksize);
3787                         goto failed_mount;
3788                 }
3789                 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3790                         le32_to_cpu(es->s_log_block_size);
3791                 sbi->s_clusters_per_group =
3792                         le32_to_cpu(es->s_clusters_per_group);
3793                 if (sbi->s_clusters_per_group > blocksize * 8) {
3794                         ext4_msg(sb, KERN_ERR,
3795                                  "#clusters per group too big: %lu",
3796                                  sbi->s_clusters_per_group);
3797                         goto failed_mount;
3798                 }
3799                 if (sbi->s_blocks_per_group !=
3800                     (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3801                         ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3802                                  "clusters per group (%lu) inconsistent",
3803                                  sbi->s_blocks_per_group,
3804                                  sbi->s_clusters_per_group);
3805                         goto failed_mount;
3806                 }
3807         } else {
3808                 if (clustersize != blocksize) {
3809                         ext4_warning(sb, "fragment/cluster size (%d) != "
3810                                      "block size (%d)", clustersize,
3811                                      blocksize);
3812                         clustersize = blocksize;
3813                 }
3814                 if (sbi->s_blocks_per_group > blocksize * 8) {
3815                         ext4_msg(sb, KERN_ERR,
3816                                  "#blocks per group too big: %lu",
3817                                  sbi->s_blocks_per_group);
3818                         goto failed_mount;
3819                 }
3820                 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3821                 sbi->s_cluster_bits = 0;
3822         }
3823         sbi->s_cluster_ratio = clustersize / blocksize;
3824
3825         if (sbi->s_inodes_per_group > blocksize * 8) {
3826                 ext4_msg(sb, KERN_ERR,
3827                        "#inodes per group too big: %lu",
3828                        sbi->s_inodes_per_group);
3829                 goto failed_mount;
3830         }
3831
3832         /* Do we have standard group size of clustersize * 8 blocks ? */
3833         if (sbi->s_blocks_per_group == clustersize << 3)
3834                 set_opt2(sb, STD_GROUP_SIZE);
3835
3836         /*
3837          * Test whether we have more sectors than will fit in sector_t,
3838          * and whether the max offset is addressable by the page cache.
3839          */
3840         err = generic_check_addressable(sb->s_blocksize_bits,
3841                                         ext4_blocks_count(es));
3842         if (err) {
3843                 ext4_msg(sb, KERN_ERR, "filesystem"
3844                          " too large to mount safely on this system");
3845                 if (sizeof(sector_t) < 8)
3846                         ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3847                 goto failed_mount;
3848         }
3849
3850         if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3851                 goto cantfind_ext4;
3852
3853         /* check blocks count against device size */
3854         blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3855         if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3856                 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3857                        "exceeds size of device (%llu blocks)",
3858                        ext4_blocks_count(es), blocks_count);
3859                 goto failed_mount;
3860         }
3861
3862         /*
3863          * It makes no sense for the first data block to be beyond the end
3864          * of the filesystem.
3865          */
3866         if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3867                 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3868                          "block %u is beyond end of filesystem (%llu)",
3869                          le32_to_cpu(es->s_first_data_block),
3870                          ext4_blocks_count(es));
3871                 goto failed_mount;
3872         }
3873         blocks_count = (ext4_blocks_count(es) -
3874                         le32_to_cpu(es->s_first_data_block) +
3875                         EXT4_BLOCKS_PER_GROUP(sb) - 1);
3876         do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3877         if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3878                 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3879                        "(block count %llu, first data block %u, "
3880                        "blocks per group %lu)", sbi->s_groups_count,
3881                        ext4_blocks_count(es),
3882                        le32_to_cpu(es->s_first_data_block),
3883                        EXT4_BLOCKS_PER_GROUP(sb));
3884                 goto failed_mount;
3885         }
3886         sbi->s_groups_count = blocks_count;
3887         sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3888                         (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3889         db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3890                    EXT4_DESC_PER_BLOCK(sb);
3891         sbi->s_group_desc = ext4_kvmalloc(db_count *
3892                                           sizeof(struct buffer_head *),
3893                                           GFP_KERNEL);
3894         if (sbi->s_group_desc == NULL) {
3895                 ext4_msg(sb, KERN_ERR, "not enough memory");
3896                 ret = -ENOMEM;
3897                 goto failed_mount;
3898         }
3899
3900         if (ext4_proc_root)
3901                 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3902
3903         if (sbi->s_proc)
3904                 proc_create_data("options", S_IRUGO, sbi->s_proc,
3905                                  &ext4_seq_options_fops, sb);
3906
3907         bgl_lock_init(sbi->s_blockgroup_lock);
3908
3909         for (i = 0; i < db_count; i++) {
3910                 block = descriptor_loc(sb, logical_sb_block, i);
3911                 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3912                 if (!sbi->s_group_desc[i]) {
3913                         ext4_msg(sb, KERN_ERR,
3914                                "can't read group descriptor %d", i);
3915                         db_count = i;
3916                         goto failed_mount2;
3917                 }
3918         }
3919         if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3920                 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3921                 goto failed_mount2;
3922         }
3923
3924         sbi->s_gdb_count = db_count;
3925         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3926         spin_lock_init(&sbi->s_next_gen_lock);
3927
3928         setup_timer(&sbi->s_err_report, print_daily_error_info,
3929                 (unsigned long) sb);
3930
3931         /* Register extent status tree shrinker */
3932         if (ext4_es_register_shrinker(sbi))
3933                 goto failed_mount3;
3934
3935         sbi->s_stripe = ext4_get_stripe_size(sbi);
3936         sbi->s_extent_max_zeroout_kb = 32;
3937
3938         /*
3939          * set up enough so that it can read an inode
3940          */
3941         sb->s_op = &ext4_sops;
3942         sb->s_export_op = &ext4_export_ops;
3943         sb->s_xattr = ext4_xattr_handlers;
3944 #ifdef CONFIG_QUOTA
3945         sb->dq_op = &ext4_quota_operations;
3946         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3947                 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3948         else
3949                 sb->s_qcop = &ext4_qctl_operations;
3950         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3951 #endif
3952         memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3953
3954         INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3955         mutex_init(&sbi->s_orphan_lock);
3956
3957         sb->s_root = NULL;
3958
3959         needs_recovery = (es->s_last_orphan != 0 ||
3960                           EXT4_HAS_INCOMPAT_FEATURE(sb,
3961                                     EXT4_FEATURE_INCOMPAT_RECOVER));
3962
3963         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3964             !(sb->s_flags & MS_RDONLY))
3965                 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3966                         goto failed_mount3a;
3967
3968         /*
3969          * The first inode we look at is the journal inode.  Don't try
3970          * root first: it may be modified in the journal!
3971          */
3972         if (!test_opt(sb, NOLOAD) &&
3973             EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3974                 if (ext4_load_journal(sb, es, journal_devnum))
3975                         goto failed_mount3a;
3976         } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3977               EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3978                 ext4_msg(sb, KERN_ERR, "required journal recovery "
3979                        "suppressed and not mounted read-only");
3980                 goto failed_mount_wq;
3981         } else {
3982                 clear_opt(sb, DATA_FLAGS);
3983                 sbi->s_journal = NULL;
3984                 needs_recovery = 0;
3985                 goto no_journal;
3986         }
3987
3988         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3989             !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3990                                        JBD2_FEATURE_INCOMPAT_64BIT)) {
3991                 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3992                 goto failed_mount_wq;
3993         }
3994
3995         if (!set_journal_csum_feature_set(sb)) {
3996                 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3997                          "feature set");
3998                 goto failed_mount_wq;
3999         }
4000
4001         /* We have now updated the journal if required, so we can
4002          * validate the data journaling mode. */
4003         switch (test_opt(sb, DATA_FLAGS)) {
4004         case 0:
4005                 /* No mode set, assume a default based on the journal
4006                  * capabilities: ORDERED_DATA if the journal can
4007                  * cope, else JOURNAL_DATA
4008                  */
4009                 if (jbd2_journal_check_available_features
4010                     (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4011                         set_opt(sb, ORDERED_DATA);
4012                 else
4013                         set_opt(sb, JOURNAL_DATA);
4014                 break;
4015
4016         case EXT4_MOUNT_ORDERED_DATA:
4017         case EXT4_MOUNT_WRITEBACK_DATA:
4018                 if (!jbd2_journal_check_available_features
4019                     (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4020                         ext4_msg(sb, KERN_ERR, "Journal does not support "
4021                                "requested data journaling mode");
4022                         goto failed_mount_wq;
4023                 }
4024         default:
4025                 break;
4026         }
4027         set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4028
4029         sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4030
4031 no_journal:
4032         if (ext4_mballoc_ready) {
4033                 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4034                 if (!sbi->s_mb_cache) {
4035                         ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4036                         goto failed_mount_wq;
4037                 }
4038         }
4039
4040         /*
4041          * Get the # of file system overhead blocks from the
4042          * superblock if present.
4043          */
4044         if (es->s_overhead_clusters)
4045                 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4046         else {
4047                 err = ext4_calculate_overhead(sb);
4048                 if (err)
4049                         goto failed_mount_wq;
4050         }
4051
4052         /*
4053          * The maximum number of concurrent works can be high and
4054          * concurrency isn't really necessary.  Limit it to 1.
4055          */
4056         EXT4_SB(sb)->rsv_conversion_wq =
4057                 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4058         if (!EXT4_SB(sb)->rsv_conversion_wq) {
4059                 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4060                 ret = -ENOMEM;
4061                 goto failed_mount4;
4062         }
4063
4064         /*
4065          * The jbd2_journal_load will have done any necessary log recovery,
4066          * so we can safely mount the rest of the filesystem now.
4067          */
4068
4069         root = ext4_iget(sb, EXT4_ROOT_INO);
4070         if (IS_ERR(root)) {
4071                 ext4_msg(sb, KERN_ERR, "get root inode failed");
4072                 ret = PTR_ERR(root);
4073                 root = NULL;
4074                 goto failed_mount4;
4075         }
4076         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4077                 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4078                 iput(root);
4079                 goto failed_mount4;
4080         }
4081         sb->s_root = d_make_root(root);
4082         if (!sb->s_root) {
4083                 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4084                 ret = -ENOMEM;
4085                 goto failed_mount4;
4086         }
4087
4088         if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4089                 sb->s_flags |= MS_RDONLY;
4090
4091         /* determine the minimum size of new large inodes, if present */
4092         if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4093                 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4094                                                      EXT4_GOOD_OLD_INODE_SIZE;
4095                 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4096                                        EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4097                         if (sbi->s_want_extra_isize <
4098                             le16_to_cpu(es->s_want_extra_isize))
4099                                 sbi->s_want_extra_isize =
4100                                         le16_to_cpu(es->s_want_extra_isize);
4101                         if (sbi->s_want_extra_isize <
4102                             le16_to_cpu(es->s_min_extra_isize))
4103                                 sbi->s_want_extra_isize =
4104                                         le16_to_cpu(es->s_min_extra_isize);
4105                 }
4106         }
4107         /* Check if enough inode space is available */
4108         if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4109                                                         sbi->s_inode_size) {
4110                 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4111                                                        EXT4_GOOD_OLD_INODE_SIZE;
4112                 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4113                          "available");
4114         }
4115
4116         err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4117         if (err) {
4118                 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4119                          "reserved pool", ext4_calculate_resv_clusters(sb));
4120                 goto failed_mount4a;
4121         }
4122
4123         err = ext4_setup_system_zone(sb);
4124         if (err) {
4125                 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4126                          "zone (%d)", err);
4127                 goto failed_mount4a;
4128         }
4129
4130         ext4_ext_init(sb);
4131         err = ext4_mb_init(sb);
4132         if (err) {
4133                 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4134                          err);
4135                 goto failed_mount5;
4136         }
4137
4138         block = ext4_count_free_clusters(sb);
4139         ext4_free_blocks_count_set(sbi->s_es,
4140                                    EXT4_C2B(sbi, block));
4141         err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4142                                   GFP_KERNEL);
4143         if (!err) {
4144                 unsigned long freei = ext4_count_free_inodes(sb);
4145                 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4146                 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4147                                           GFP_KERNEL);
4148         }
4149         if (!err)
4150                 err = percpu_counter_init(&sbi->s_dirs_counter,
4151                                           ext4_count_dirs(sb), GFP_KERNEL);
4152         if (!err)
4153                 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4154                                           GFP_KERNEL);
4155         if (err) {
4156                 ext4_msg(sb, KERN_ERR, "insufficient memory");
4157                 goto failed_mount6;
4158         }
4159
4160         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4161                 if (!ext4_fill_flex_info(sb)) {
4162                         ext4_msg(sb, KERN_ERR,
4163                                "unable to initialize "
4164                                "flex_bg meta info!");
4165                         goto failed_mount6;
4166                 }
4167
4168         err = ext4_register_li_request(sb, first_not_zeroed);
4169         if (err)
4170                 goto failed_mount6;
4171
4172         sbi->s_kobj.kset = ext4_kset;
4173         init_completion(&sbi->s_kobj_unregister);
4174         err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4175                                    "%s", sb->s_id);
4176         if (err)
4177                 goto failed_mount7;
4178
4179 #ifdef CONFIG_QUOTA
4180         /* Enable quota usage during mount. */
4181         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4182             !(sb->s_flags & MS_RDONLY)) {
4183                 err = ext4_enable_quotas(sb);
4184                 if (err)
4185                         goto failed_mount8;
4186         }
4187 #endif  /* CONFIG_QUOTA */
4188
4189         EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4190         ext4_orphan_cleanup(sb, es);
4191         EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4192         if (needs_recovery) {
4193                 ext4_msg(sb, KERN_INFO, "recovery complete");
4194                 ext4_mark_recovery_complete(sb, es);
4195         }
4196         if (EXT4_SB(sb)->s_journal) {
4197                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4198                         descr = " journalled data mode";
4199                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4200                         descr = " ordered data mode";
4201                 else
4202                         descr = " writeback data mode";
4203         } else
4204                 descr = "out journal";
4205
4206         if (test_opt(sb, DISCARD)) {
4207                 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4208                 if (!blk_queue_discard(q))
4209                         ext4_msg(sb, KERN_WARNING,
4210                                  "mounting with \"discard\" option, but "
4211                                  "the device does not support discard");
4212         }
4213
4214         ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4215                  "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4216                  *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4217
4218         if (es->s_error_count)
4219                 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4220
4221         /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4222         ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4223         ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4224         ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4225
4226         kfree(orig_data);
4227         return 0;
4228
4229 cantfind_ext4:
4230         if (!silent)
4231                 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4232         goto failed_mount;
4233
4234 #ifdef CONFIG_QUOTA
4235 failed_mount8:
4236         kobject_del(&sbi->s_kobj);
4237 #endif
4238 failed_mount7:
4239         ext4_unregister_li_request(sb);
4240 failed_mount6:
4241         ext4_mb_release(sb);
4242         if (sbi->s_flex_groups)
4243                 kvfree(sbi->s_flex_groups);
4244         percpu_counter_destroy(&sbi->s_freeclusters_counter);
4245         percpu_counter_destroy(&sbi->s_freeinodes_counter);
4246         percpu_counter_destroy(&sbi->s_dirs_counter);
4247         percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4248 failed_mount5:
4249         ext4_ext_release(sb);
4250         ext4_release_system_zone(sb);
4251 failed_mount4a:
4252         dput(sb->s_root);
4253         sb->s_root = NULL;
4254 failed_mount4:
4255         ext4_msg(sb, KERN_ERR, "mount failed");
4256         if (EXT4_SB(sb)->rsv_conversion_wq)
4257                 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4258 failed_mount_wq:
4259         if (sbi->s_journal) {
4260                 jbd2_journal_destroy(sbi->s_journal);
4261                 sbi->s_journal = NULL;
4262         }
4263 failed_mount3a:
4264         ext4_es_unregister_shrinker(sbi);
4265 failed_mount3:
4266         del_timer_sync(&sbi->s_err_report);
4267         if (sbi->s_mmp_tsk)
4268                 kthread_stop(sbi->s_mmp_tsk);
4269 failed_mount2:
4270         for (i = 0; i < db_count; i++)
4271                 brelse(sbi->s_group_desc[i]);
4272         kvfree(sbi->s_group_desc);
4273 failed_mount:
4274         if (sbi->s_chksum_driver)
4275                 crypto_free_shash(sbi->s_chksum_driver);
4276         if (sbi->s_proc) {
4277                 remove_proc_entry("options", sbi->s_proc);
4278                 remove_proc_entry(sb->s_id, ext4_proc_root);
4279         }
4280 #ifdef CONFIG_QUOTA
4281         for (i = 0; i < EXT4_MAXQUOTAS; i++)
4282                 kfree(sbi->s_qf_names[i]);
4283 #endif
4284         ext4_blkdev_remove(sbi);
4285         brelse(bh);
4286 out_fail:
4287         sb->s_fs_info = NULL;
4288         kfree(sbi->s_blockgroup_lock);
4289         kfree(sbi);
4290 out_free_orig:
4291         kfree(orig_data);
4292         return err ? err : ret;
4293 }
4294
4295 /*
4296  * Setup any per-fs journal parameters now.  We'll do this both on
4297  * initial mount, once the journal has been initialised but before we've
4298  * done any recovery; and again on any subsequent remount.
4299  */
4300 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4301 {
4302         struct ext4_sb_info *sbi = EXT4_SB(sb);
4303
4304         journal->j_commit_interval = sbi->s_commit_interval;
4305         journal->j_min_batch_time = sbi->s_min_batch_time;
4306         journal->j_max_batch_time = sbi->s_max_batch_time;
4307
4308         write_lock(&journal->j_state_lock);
4309         if (test_opt(sb, BARRIER))
4310                 journal->j_flags |= JBD2_BARRIER;
4311         else
4312                 journal->j_flags &= ~JBD2_BARRIER;
4313         if (test_opt(sb, DATA_ERR_ABORT))
4314                 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4315         else
4316                 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4317         write_unlock(&journal->j_state_lock);
4318 }
4319
4320 static journal_t *ext4_get_journal(struct super_block *sb,
4321                                    unsigned int journal_inum)
4322 {
4323         struct inode *journal_inode;
4324         journal_t *journal;
4325
4326         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4327
4328         /* First, test for the existence of a valid inode on disk.  Bad
4329          * things happen if we iget() an unused inode, as the subsequent
4330          * iput() will try to delete it. */
4331
4332         journal_inode = ext4_iget(sb, journal_inum);
4333         if (IS_ERR(journal_inode)) {
4334                 ext4_msg(sb, KERN_ERR, "no journal found");
4335                 return NULL;
4336         }
4337         if (!journal_inode->i_nlink) {
4338                 make_bad_inode(journal_inode);
4339                 iput(journal_inode);
4340                 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4341                 return NULL;
4342         }
4343
4344         jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4345                   journal_inode, journal_inode->i_size);
4346         if (!S_ISREG(journal_inode->i_mode)) {
4347                 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4348                 iput(journal_inode);
4349                 return NULL;
4350         }
4351
4352         journal = jbd2_journal_init_inode(journal_inode);
4353         if (!journal) {
4354                 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4355                 iput(journal_inode);
4356                 return NULL;
4357         }
4358         journal->j_private = sb;
4359         ext4_init_journal_params(sb, journal);
4360         return journal;
4361 }
4362
4363 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4364                                        dev_t j_dev)
4365 {
4366         struct buffer_head *bh;
4367         journal_t *journal;
4368         ext4_fsblk_t start;
4369         ext4_fsblk_t len;
4370         int hblock, blocksize;
4371         ext4_fsblk_t sb_block;
4372         unsigned long offset;
4373         struct ext4_super_block *es;
4374         struct block_device *bdev;
4375
4376         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4377
4378         bdev = ext4_blkdev_get(j_dev, sb);
4379         if (bdev == NULL)
4380                 return NULL;
4381
4382         blocksize = sb->s_blocksize;
4383         hblock = bdev_logical_block_size(bdev);
4384         if (blocksize < hblock) {
4385                 ext4_msg(sb, KERN_ERR,
4386                         "blocksize too small for journal device");
4387                 goto out_bdev;
4388         }
4389
4390         sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4391         offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4392         set_blocksize(bdev, blocksize);
4393         if (!(bh = __bread(bdev, sb_block, blocksize))) {
4394                 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4395                        "external journal");
4396                 goto out_bdev;
4397         }
4398
4399         es = (struct ext4_super_block *) (bh->b_data + offset);
4400         if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4401             !(le32_to_cpu(es->s_feature_incompat) &
4402               EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4403                 ext4_msg(sb, KERN_ERR, "external journal has "
4404                                         "bad superblock");
4405                 brelse(bh);
4406                 goto out_bdev;
4407         }
4408
4409         if ((le32_to_cpu(es->s_feature_ro_compat) &
4410              EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4411             es->s_checksum != ext4_superblock_csum(sb, es)) {
4412                 ext4_msg(sb, KERN_ERR, "external journal has "
4413                                        "corrupt superblock");
4414                 brelse(bh);
4415                 goto out_bdev;
4416         }
4417
4418         if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4419                 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4420                 brelse(bh);
4421                 goto out_bdev;
4422         }
4423
4424         len = ext4_blocks_count(es);
4425         start = sb_block + 1;
4426         brelse(bh);     /* we're done with the superblock */
4427
4428         journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4429                                         start, len, blocksize);
4430         if (!journal) {
4431                 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4432                 goto out_bdev;
4433         }
4434         journal->j_private = sb;
4435         ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4436         wait_on_buffer(journal->j_sb_buffer);
4437         if (!buffer_uptodate(journal->j_sb_buffer)) {
4438                 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4439                 goto out_journal;
4440         }
4441         if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4442                 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4443                                         "user (unsupported) - %d",
4444                         be32_to_cpu(journal->j_superblock->s_nr_users));
4445                 goto out_journal;
4446         }
4447         EXT4_SB(sb)->journal_bdev = bdev;
4448         ext4_init_journal_params(sb, journal);
4449         return journal;
4450
4451 out_journal:
4452         jbd2_journal_destroy(journal);
4453 out_bdev:
4454         ext4_blkdev_put(bdev);
4455         return NULL;
4456 }
4457
4458 static int ext4_load_journal(struct super_block *sb,
4459                              struct ext4_super_block *es,
4460                              unsigned long journal_devnum)
4461 {
4462         journal_t *journal;
4463         unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4464         dev_t journal_dev;
4465         int err = 0;
4466         int really_read_only;
4467
4468         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4469
4470         if (journal_devnum &&
4471             journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4472                 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4473                         "numbers have changed");
4474                 journal_dev = new_decode_dev(journal_devnum);
4475         } else
4476                 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4477
4478         really_read_only = bdev_read_only(sb->s_bdev);
4479
4480         /*
4481          * Are we loading a blank journal or performing recovery after a
4482          * crash?  For recovery, we need to check in advance whether we
4483          * can get read-write access to the device.
4484          */
4485         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4486                 if (sb->s_flags & MS_RDONLY) {
4487                         ext4_msg(sb, KERN_INFO, "INFO: recovery "
4488                                         "required on readonly filesystem");
4489                         if (really_read_only) {
4490                                 ext4_msg(sb, KERN_ERR, "write access "
4491                                         "unavailable, cannot proceed");
4492                                 return -EROFS;
4493                         }
4494                         ext4_msg(sb, KERN_INFO, "write access will "
4495                                "be enabled during recovery");
4496                 }
4497         }
4498
4499         if (journal_inum && journal_dev) {
4500                 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4501                        "and inode journals!");
4502                 return -EINVAL;
4503         }
4504
4505         if (journal_inum) {
4506                 if (!(journal = ext4_get_journal(sb, journal_inum)))
4507                         return -EINVAL;
4508         } else {
4509                 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4510                         return -EINVAL;
4511         }
4512
4513         if (!(journal->j_flags & JBD2_BARRIER))
4514                 ext4_msg(sb, KERN_INFO, "barriers disabled");
4515
4516         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4517                 err = jbd2_journal_wipe(journal, !really_read_only);
4518         if (!err) {
4519                 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4520                 if (save)
4521                         memcpy(save, ((char *) es) +
4522                                EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4523                 err = jbd2_journal_load(journal);
4524                 if (save)
4525                         memcpy(((char *) es) + EXT4_S_ERR_START,
4526                                save, EXT4_S_ERR_LEN);
4527                 kfree(save);
4528         }
4529
4530         if (err) {
4531                 ext4_msg(sb, KERN_ERR, "error loading journal");
4532                 jbd2_journal_destroy(journal);
4533                 return err;
4534         }
4535
4536         EXT4_SB(sb)->s_journal = journal;
4537         ext4_clear_journal_err(sb, es);
4538
4539         if (!really_read_only && journal_devnum &&
4540             journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4541                 es->s_journal_dev = cpu_to_le32(journal_devnum);
4542
4543                 /* Make sure we flush the recovery flag to disk. */
4544                 ext4_commit_super(sb, 1);
4545         }
4546
4547         return 0;
4548 }
4549
4550 static int ext4_commit_super(struct super_block *sb, int sync)
4551 {
4552         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4553         struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4554         int error = 0;
4555
4556         if (!sbh)
4557                 return error;
4558         if (buffer_write_io_error(sbh)) {
4559                 /*
4560                  * Oh, dear.  A previous attempt to write the
4561                  * superblock failed.  This could happen because the
4562                  * USB device was yanked out.  Or it could happen to
4563                  * be a transient write error and maybe the block will
4564                  * be remapped.  Nothing we can do but to retry the
4565                  * write and hope for the best.
4566                  */
4567                 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4568                        "superblock detected");
4569                 clear_buffer_write_io_error(sbh);
4570                 set_buffer_uptodate(sbh);
4571         }
4572         /*
4573          * If the file system is mounted read-only, don't update the
4574          * superblock write time.  This avoids updating the superblock
4575          * write time when we are mounting the root file system
4576          * read/only but we need to replay the journal; at that point,
4577          * for people who are east of GMT and who make their clock
4578          * tick in localtime for Windows bug-for-bug compatibility,
4579          * the clock is set in the future, and this will cause e2fsck
4580          * to complain and force a full file system check.
4581          */
4582         if (!(sb->s_flags & MS_RDONLY))
4583                 es->s_wtime = cpu_to_le32(get_seconds());
4584         if (sb->s_bdev->bd_part)
4585                 es->s_kbytes_written =
4586                         cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4587                             ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4588                               EXT4_SB(sb)->s_sectors_written_start) >> 1));
4589         else
4590                 es->s_kbytes_written =
4591                         cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4592         if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4593                 ext4_free_blocks_count_set(es,
4594                         EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4595                                 &EXT4_SB(sb)->s_freeclusters_counter)));
4596         if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4597                 es->s_free_inodes_count =
4598                         cpu_to_le32(percpu_counter_sum_positive(
4599                                 &EXT4_SB(sb)->s_freeinodes_counter));
4600         BUFFER_TRACE(sbh, "marking dirty");
4601         ext4_superblock_csum_set(sb);
4602         mark_buffer_dirty(sbh);
4603         if (sync) {
4604                 error = sync_dirty_buffer(sbh);
4605                 if (error)
4606                         return error;
4607
4608                 error = buffer_write_io_error(sbh);
4609                 if (error) {
4610                         ext4_msg(sb, KERN_ERR, "I/O error while writing "
4611                                "superblock");
4612                         clear_buffer_write_io_error(sbh);
4613                         set_buffer_uptodate(sbh);
4614                 }
4615         }
4616         return error;
4617 }
4618
4619 /*
4620  * Have we just finished recovery?  If so, and if we are mounting (or
4621  * remounting) the filesystem readonly, then we will end up with a
4622  * consistent fs on disk.  Record that fact.
4623  */
4624 static void ext4_mark_recovery_complete(struct super_block *sb,
4625                                         struct ext4_super_block *es)
4626 {
4627         journal_t *journal = EXT4_SB(sb)->s_journal;
4628
4629         if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4630                 BUG_ON(journal != NULL);
4631                 return;
4632         }
4633         jbd2_journal_lock_updates(journal);
4634         if (jbd2_journal_flush(journal) < 0)
4635                 goto out;
4636
4637         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4638             sb->s_flags & MS_RDONLY) {
4639                 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4640                 ext4_commit_super(sb, 1);
4641         }
4642
4643 out:
4644         jbd2_journal_unlock_updates(journal);
4645 }
4646
4647 /*
4648  * If we are mounting (or read-write remounting) a filesystem whose journal
4649  * has recorded an error from a previous lifetime, move that error to the
4650  * main filesystem now.
4651  */
4652 static void ext4_clear_journal_err(struct super_block *sb,
4653                                    struct ext4_super_block *es)
4654 {
4655         journal_t *journal;
4656         int j_errno;
4657         const char *errstr;
4658
4659         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4660
4661         journal = EXT4_SB(sb)->s_journal;
4662
4663         /*
4664          * Now check for any error status which may have been recorded in the
4665          * journal by a prior ext4_error() or ext4_abort()
4666          */
4667
4668         j_errno = jbd2_journal_errno(journal);
4669         if (j_errno) {
4670                 char nbuf[16];
4671
4672                 errstr = ext4_decode_error(sb, j_errno, nbuf);
4673                 ext4_warning(sb, "Filesystem error recorded "
4674                              "from previous mount: %s", errstr);
4675                 ext4_warning(sb, "Marking fs in need of filesystem check.");
4676
4677                 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4678                 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4679                 ext4_commit_super(sb, 1);
4680
4681                 jbd2_journal_clear_err(journal);
4682                 jbd2_journal_update_sb_errno(journal);
4683         }
4684 }
4685
4686 /*
4687  * Force the running and committing transactions to commit,
4688  * and wait on the commit.
4689  */
4690 int ext4_force_commit(struct super_block *sb)
4691 {
4692         journal_t *journal;
4693
4694         if (sb->s_flags & MS_RDONLY)
4695                 return 0;
4696
4697         journal = EXT4_SB(sb)->s_journal;
4698         return ext4_journal_force_commit(journal);
4699 }
4700
4701 static int ext4_sync_fs(struct super_block *sb, int wait)
4702 {
4703         int ret = 0;
4704         tid_t target;
4705         bool needs_barrier = false;
4706         struct ext4_sb_info *sbi = EXT4_SB(sb);
4707
4708         trace_ext4_sync_fs(sb, wait);
4709         flush_workqueue(sbi->rsv_conversion_wq);
4710         /*
4711          * Writeback quota in non-journalled quota case - journalled quota has
4712          * no dirty dquots
4713          */
4714         dquot_writeback_dquots(sb, -1);
4715         /*
4716          * Data writeback is possible w/o journal transaction, so barrier must
4717          * being sent at the end of the function. But we can skip it if
4718          * transaction_commit will do it for us.
4719          */
4720         if (sbi->s_journal) {
4721                 target = jbd2_get_latest_transaction(sbi->s_journal);
4722                 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4723                     !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4724                         needs_barrier = true;
4725
4726                 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4727                         if (wait)
4728                                 ret = jbd2_log_wait_commit(sbi->s_journal,
4729                                                            target);
4730                 }
4731         } else if (wait && test_opt(sb, BARRIER))
4732                 needs_barrier = true;
4733         if (needs_barrier) {
4734                 int err;
4735                 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4736                 if (!ret)
4737                         ret = err;
4738         }
4739
4740         return ret;
4741 }
4742
4743 /*
4744  * LVM calls this function before a (read-only) snapshot is created.  This
4745  * gives us a chance to flush the journal completely and mark the fs clean.
4746  *
4747  * Note that only this function cannot bring a filesystem to be in a clean
4748  * state independently. It relies on upper layer to stop all data & metadata
4749  * modifications.
4750  */
4751 static int ext4_freeze(struct super_block *sb)
4752 {
4753         int error = 0;
4754         journal_t *journal;
4755
4756         if (sb->s_flags & MS_RDONLY)
4757                 return 0;
4758
4759         journal = EXT4_SB(sb)->s_journal;
4760
4761         if (journal) {
4762                 /* Now we set up the journal barrier. */
4763                 jbd2_journal_lock_updates(journal);
4764
4765                 /*
4766                  * Don't clear the needs_recovery flag if we failed to
4767                  * flush the journal.
4768                  */
4769                 error = jbd2_journal_flush(journal);
4770                 if (error < 0)
4771                         goto out;
4772         }
4773
4774         /* Journal blocked and flushed, clear needs_recovery flag. */
4775         EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4776         error = ext4_commit_super(sb, 1);
4777 out:
4778         if (journal)
4779                 /* we rely on upper layer to stop further updates */
4780                 jbd2_journal_unlock_updates(journal);
4781         return error;
4782 }
4783
4784 /*
4785  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4786  * flag here, even though the filesystem is not technically dirty yet.
4787  */
4788 static int ext4_unfreeze(struct super_block *sb)
4789 {
4790         if (sb->s_flags & MS_RDONLY)
4791                 return 0;
4792
4793         /* Reset the needs_recovery flag before the fs is unlocked. */
4794         EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4795         ext4_commit_super(sb, 1);
4796         return 0;
4797 }
4798
4799 /*
4800  * Structure to save mount options for ext4_remount's benefit
4801  */
4802 struct ext4_mount_options {
4803         unsigned long s_mount_opt;
4804         unsigned long s_mount_opt2;
4805         kuid_t s_resuid;
4806         kgid_t s_resgid;
4807         unsigned long s_commit_interval;
4808         u32 s_min_batch_time, s_max_batch_time;
4809 #ifdef CONFIG_QUOTA
4810         int s_jquota_fmt;
4811         char *s_qf_names[EXT4_MAXQUOTAS];
4812 #endif
4813 };
4814
4815 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4816 {
4817         struct ext4_super_block *es;
4818         struct ext4_sb_info *sbi = EXT4_SB(sb);
4819         unsigned long old_sb_flags;
4820         struct ext4_mount_options old_opts;
4821         int enable_quota = 0;
4822         ext4_group_t g;
4823         unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4824         int err = 0;
4825 #ifdef CONFIG_QUOTA
4826         int i, j;
4827 #endif
4828         char *orig_data = kstrdup(data, GFP_KERNEL);
4829
4830         /* Store the original options */
4831         old_sb_flags = sb->s_flags;
4832         old_opts.s_mount_opt = sbi->s_mount_opt;
4833         old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4834         old_opts.s_resuid = sbi->s_resuid;
4835         old_opts.s_resgid = sbi->s_resgid;
4836         old_opts.s_commit_interval = sbi->s_commit_interval;
4837         old_opts.s_min_batch_time = sbi->s_min_batch_time;
4838         old_opts.s_max_batch_time = sbi->s_max_batch_time;
4839 #ifdef CONFIG_QUOTA
4840         old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4841         for (i = 0; i < EXT4_MAXQUOTAS; i++)
4842                 if (sbi->s_qf_names[i]) {
4843                         old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4844                                                          GFP_KERNEL);
4845                         if (!old_opts.s_qf_names[i]) {
4846                                 for (j = 0; j < i; j++)
4847                                         kfree(old_opts.s_qf_names[j]);
4848                                 kfree(orig_data);
4849                                 return -ENOMEM;
4850                         }
4851                 } else
4852                         old_opts.s_qf_names[i] = NULL;
4853 #endif
4854         if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4855                 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4856
4857         if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4858                 err = -EINVAL;
4859                 goto restore_opts;
4860         }
4861
4862         if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4863             test_opt(sb, JOURNAL_CHECKSUM)) {
4864                 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4865                          "during remount not supported; ignoring");
4866                 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4867         }
4868
4869         if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4870                 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4871                         ext4_msg(sb, KERN_ERR, "can't mount with "
4872                                  "both data=journal and delalloc");
4873                         err = -EINVAL;
4874                         goto restore_opts;
4875                 }
4876                 if (test_opt(sb, DIOREAD_NOLOCK)) {
4877                         ext4_msg(sb, KERN_ERR, "can't mount with "
4878                                  "both data=journal and dioread_nolock");
4879                         err = -EINVAL;
4880                         goto restore_opts;
4881                 }
4882                 if (test_opt(sb, DAX)) {
4883                         ext4_msg(sb, KERN_ERR, "can't mount with "
4884                                  "both data=journal and dax");
4885                         err = -EINVAL;
4886                         goto restore_opts;
4887                 }
4888         }
4889
4890         if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4891                 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4892                         "dax flag with busy inodes while remounting");
4893                 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4894         }
4895
4896         if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4897                 ext4_abort(sb, "Abort forced by user");
4898
4899         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4900                 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4901
4902         es = sbi->s_es;
4903
4904         if (sbi->s_journal) {
4905                 ext4_init_journal_params(sb, sbi->s_journal);
4906                 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4907         }
4908
4909         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4910                 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4911                         err = -EROFS;
4912                         goto restore_opts;
4913                 }
4914
4915                 if (*flags & MS_RDONLY) {
4916                         err = sync_filesystem(sb);
4917                         if (err < 0)
4918                                 goto restore_opts;
4919                         err = dquot_suspend(sb, -1);
4920                         if (err < 0)
4921                                 goto restore_opts;
4922
4923                         /*
4924                          * First of all, the unconditional stuff we have to do
4925                          * to disable replay of the journal when we next remount
4926                          */
4927                         sb->s_flags |= MS_RDONLY;
4928
4929                         /*
4930                          * OK, test if we are remounting a valid rw partition
4931                          * readonly, and if so set the rdonly flag and then
4932                          * mark the partition as valid again.
4933                          */
4934                         if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4935                             (sbi->s_mount_state & EXT4_VALID_FS))
4936                                 es->s_state = cpu_to_le16(sbi->s_mount_state);
4937
4938                         if (sbi->s_journal)
4939                                 ext4_mark_recovery_complete(sb, es);
4940                 } else {
4941                         /* Make sure we can mount this feature set readwrite */
4942                         if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4943                                         EXT4_FEATURE_RO_COMPAT_READONLY) ||
4944                             !ext4_feature_set_ok(sb, 0)) {
4945                                 err = -EROFS;
4946                                 goto restore_opts;
4947                         }
4948                         /*
4949                          * Make sure the group descriptor checksums
4950                          * are sane.  If they aren't, refuse to remount r/w.
4951                          */
4952                         for (g = 0; g < sbi->s_groups_count; g++) {
4953                                 struct ext4_group_desc *gdp =
4954                                         ext4_get_group_desc(sb, g, NULL);
4955
4956                                 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4957                                         ext4_msg(sb, KERN_ERR,
4958                "ext4_remount: Checksum for group %u failed (%u!=%u)",
4959                 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4960                                                le16_to_cpu(gdp->bg_checksum));
4961                                         err = -EINVAL;
4962                                         goto restore_opts;
4963                                 }
4964                         }
4965
4966                         /*
4967                          * If we have an unprocessed orphan list hanging
4968                          * around from a previously readonly bdev mount,
4969                          * require a full umount/remount for now.
4970                          */
4971                         if (es->s_last_orphan) {
4972                                 ext4_msg(sb, KERN_WARNING, "Couldn't "
4973                                        "remount RDWR because of unprocessed "
4974                                        "orphan inode list.  Please "
4975                                        "umount/remount instead");
4976                                 err = -EINVAL;
4977                                 goto restore_opts;
4978                         }
4979
4980                         /*
4981                          * Mounting a RDONLY partition read-write, so reread
4982                          * and store the current valid flag.  (It may have
4983                          * been changed by e2fsck since we originally mounted
4984                          * the partition.)
4985                          */
4986                         if (sbi->s_journal)
4987                                 ext4_clear_journal_err(sb, es);
4988                         sbi->s_mount_state = le16_to_cpu(es->s_state);
4989                         if (!ext4_setup_super(sb, es, 0))
4990                                 sb->s_flags &= ~MS_RDONLY;
4991                         if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4992                                                      EXT4_FEATURE_INCOMPAT_MMP))
4993                                 if (ext4_multi_mount_protect(sb,
4994                                                 le64_to_cpu(es->s_mmp_block))) {
4995                                         err = -EROFS;
4996                                         goto restore_opts;
4997                                 }
4998                         enable_quota = 1;
4999                 }
5000         }
5001
5002         /*
5003          * Reinitialize lazy itable initialization thread based on
5004          * current settings
5005          */
5006         if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5007                 ext4_unregister_li_request(sb);
5008         else {
5009                 ext4_group_t first_not_zeroed;
5010                 first_not_zeroed = ext4_has_uninit_itable(sb);
5011                 ext4_register_li_request(sb, first_not_zeroed);
5012         }
5013
5014         ext4_setup_system_zone(sb);
5015         if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5016                 ext4_commit_super(sb, 1);
5017
5018 #ifdef CONFIG_QUOTA
5019         /* Release old quota file names */
5020         for (i = 0; i < EXT4_MAXQUOTAS; i++)
5021                 kfree(old_opts.s_qf_names[i]);
5022         if (enable_quota) {
5023                 if (sb_any_quota_suspended(sb))
5024                         dquot_resume(sb, -1);
5025                 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5026                                         EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5027                         err = ext4_enable_quotas(sb);
5028                         if (err)
5029                                 goto restore_opts;
5030                 }
5031         }
5032 #endif
5033
5034         *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5035         ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5036         kfree(orig_data);
5037         return 0;
5038
5039 restore_opts:
5040         sb->s_flags = old_sb_flags;
5041         sbi->s_mount_opt = old_opts.s_mount_opt;
5042         sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5043         sbi->s_resuid = old_opts.s_resuid;
5044         sbi->s_resgid = old_opts.s_resgid;
5045         sbi->s_commit_interval = old_opts.s_commit_interval;
5046         sbi->s_min_batch_time = old_opts.s_min_batch_time;
5047         sbi->s_max_batch_time = old_opts.s_max_batch_time;
5048 #ifdef CONFIG_QUOTA
5049         sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5050         for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5051                 kfree(sbi->s_qf_names[i]);
5052                 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5053         }
5054 #endif
5055         kfree(orig_data);
5056         return err;
5057 }
5058
5059 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5060 {
5061         struct super_block *sb = dentry->d_sb;
5062         struct ext4_sb_info *sbi = EXT4_SB(sb);
5063         struct ext4_super_block *es = sbi->s_es;
5064         ext4_fsblk_t overhead = 0, resv_blocks;
5065         u64 fsid;
5066         s64 bfree;
5067         resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5068
5069         if (!test_opt(sb, MINIX_DF))
5070                 overhead = sbi->s_overhead;
5071
5072         buf->f_type = EXT4_SUPER_MAGIC;
5073         buf->f_bsize = sb->s_blocksize;
5074         buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5075         bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5076                 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5077         /* prevent underflow in case that few free space is available */
5078         buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5079         buf->f_bavail = buf->f_bfree -
5080                         (ext4_r_blocks_count(es) + resv_blocks);
5081         if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5082                 buf->f_bavail = 0;
5083         buf->f_files = le32_to_cpu(es->s_inodes_count);
5084         buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5085         buf->f_namelen = EXT4_NAME_LEN;
5086         fsid = le64_to_cpup((void *)es->s_uuid) ^
5087                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5088         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5089         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5090
5091         return 0;
5092 }
5093
5094 /* Helper function for writing quotas on sync - we need to start transaction
5095  * before quota file is locked for write. Otherwise the are possible deadlocks:
5096  * Process 1                         Process 2
5097  * ext4_create()                     quota_sync()
5098  *   jbd2_journal_start()                  write_dquot()
5099  *   dquot_initialize()                         down(dqio_mutex)
5100  *     down(dqio_mutex)                    jbd2_journal_start()
5101  *
5102  */
5103
5104 #ifdef CONFIG_QUOTA
5105
5106 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5107 {
5108         return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5109 }
5110
5111 static int ext4_write_dquot(struct dquot *dquot)
5112 {
5113         int ret, err;
5114         handle_t *handle;
5115         struct inode *inode;
5116
5117         inode = dquot_to_inode(dquot);
5118         handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5119                                     EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5120         if (IS_ERR(handle))
5121                 return PTR_ERR(handle);
5122         ret = dquot_commit(dquot);
5123         err = ext4_journal_stop(handle);
5124         if (!ret)
5125                 ret = err;
5126         return ret;
5127 }
5128
5129 static int ext4_acquire_dquot(struct dquot *dquot)
5130 {
5131         int ret, err;
5132         handle_t *handle;
5133
5134         handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5135                                     EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5136         if (IS_ERR(handle))
5137                 return PTR_ERR(handle);
5138         ret = dquot_acquire(dquot);
5139         err = ext4_journal_stop(handle);
5140         if (!ret)
5141                 ret = err;
5142         return ret;
5143 }
5144
5145 static int ext4_release_dquot(struct dquot *dquot)
5146 {
5147         int ret, err;
5148         handle_t *handle;
5149
5150         handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5151                                     EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5152         if (IS_ERR(handle)) {
5153                 /* Release dquot anyway to avoid endless cycle in dqput() */
5154                 dquot_release(dquot);
5155                 return PTR_ERR(handle);
5156         }
5157         ret = dquot_release(dquot);
5158         err = ext4_journal_stop(handle);
5159         if (!ret)
5160                 ret = err;
5161         return ret;
5162 }
5163
5164 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5165 {
5166         struct super_block *sb = dquot->dq_sb;
5167         struct ext4_sb_info *sbi = EXT4_SB(sb);
5168
5169         /* Are we journaling quotas? */
5170         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5171             sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5172                 dquot_mark_dquot_dirty(dquot);
5173                 return ext4_write_dquot(dquot);
5174         } else {
5175                 return dquot_mark_dquot_dirty(dquot);
5176         }
5177 }
5178
5179 static int ext4_write_info(struct super_block *sb, int type)
5180 {
5181         int ret, err;
5182         handle_t *handle;
5183
5184         /* Data block + inode block */
5185         handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5186         if (IS_ERR(handle))
5187                 return PTR_ERR(handle);
5188         ret = dquot_commit_info(sb, type);
5189         err = ext4_journal_stop(handle);
5190         if (!ret)
5191                 ret = err;
5192         return ret;
5193 }
5194
5195 /*
5196  * Turn on quotas during mount time - we need to find
5197  * the quota file and such...
5198  */
5199 static int ext4_quota_on_mount(struct super_block *sb, int type)
5200 {
5201         return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5202                                         EXT4_SB(sb)->s_jquota_fmt, type);
5203 }
5204
5205 /*
5206  * Standard function to be called on quota_on
5207  */
5208 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5209                          struct path *path)
5210 {
5211         int err;
5212
5213         if (!test_opt(sb, QUOTA))
5214                 return -EINVAL;
5215
5216         /* Quotafile not on the same filesystem? */
5217         if (path->dentry->d_sb != sb)
5218                 return -EXDEV;
5219         /* Journaling quota? */
5220         if (EXT4_SB(sb)->s_qf_names[type]) {
5221                 /* Quotafile not in fs root? */
5222                 if (path->dentry->d_parent != sb->s_root)
5223                         ext4_msg(sb, KERN_WARNING,
5224                                 "Quota file not on filesystem root. "
5225                                 "Journaled quota will not work");
5226         }
5227
5228         /*
5229          * When we journal data on quota file, we have to flush journal to see
5230          * all updates to the file when we bypass pagecache...
5231          */
5232         if (EXT4_SB(sb)->s_journal &&
5233             ext4_should_journal_data(path->dentry->d_inode)) {
5234                 /*
5235                  * We don't need to lock updates but journal_flush() could
5236                  * otherwise be livelocked...
5237                  */
5238                 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5239                 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5240                 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5241                 if (err)
5242                         return err;
5243         }
5244
5245         return dquot_quota_on(sb, type, format_id, path);
5246 }
5247
5248 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5249                              unsigned int flags)
5250 {
5251         int err;
5252         struct inode *qf_inode;
5253         unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5254                 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5255                 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5256         };
5257
5258         BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5259
5260         if (!qf_inums[type])
5261                 return -EPERM;
5262
5263         qf_inode = ext4_iget(sb, qf_inums[type]);
5264         if (IS_ERR(qf_inode)) {
5265                 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5266                 return PTR_ERR(qf_inode);
5267         }
5268
5269         /* Don't account quota for quota files to avoid recursion */
5270         qf_inode->i_flags |= S_NOQUOTA;
5271         err = dquot_enable(qf_inode, type, format_id, flags);
5272         iput(qf_inode);
5273
5274         return err;
5275 }
5276
5277 /* Enable usage tracking for all quota types. */
5278 static int ext4_enable_quotas(struct super_block *sb)
5279 {
5280         int type, err = 0;
5281         unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5282                 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5283                 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5284         };
5285
5286         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5287         for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5288                 if (qf_inums[type]) {
5289                         err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5290                                                 DQUOT_USAGE_ENABLED);
5291                         if (err) {
5292                                 ext4_warning(sb,
5293                                         "Failed to enable quota tracking "
5294                                         "(type=%d, err=%d). Please run "
5295                                         "e2fsck to fix.", type, err);
5296                                 return err;
5297                         }
5298                 }
5299         }
5300         return 0;
5301 }
5302
5303 static int ext4_quota_off(struct super_block *sb, int type)
5304 {
5305         struct inode *inode = sb_dqopt(sb)->files[type];
5306         handle_t *handle;
5307
5308         /* Force all delayed allocation blocks to be allocated.
5309          * Caller already holds s_umount sem */
5310         if (test_opt(sb, DELALLOC))
5311                 sync_filesystem(sb);
5312
5313         if (!inode)
5314                 goto out;
5315
5316         /* Update modification times of quota files when userspace can
5317          * start looking at them */
5318         handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5319         if (IS_ERR(handle))
5320                 goto out;
5321         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5322         ext4_mark_inode_dirty(handle, inode);
5323         ext4_journal_stop(handle);
5324
5325 out:
5326         return dquot_quota_off(sb, type);
5327 }
5328
5329 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5330  * acquiring the locks... As quota files are never truncated and quota code
5331  * itself serializes the operations (and no one else should touch the files)
5332  * we don't have to be afraid of races */
5333 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5334                                size_t len, loff_t off)
5335 {
5336         struct inode *inode = sb_dqopt(sb)->files[type];
5337         ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5338         int offset = off & (sb->s_blocksize - 1);
5339         int tocopy;
5340         size_t toread;
5341         struct buffer_head *bh;
5342         loff_t i_size = i_size_read(inode);
5343
5344         if (off > i_size)
5345                 return 0;
5346         if (off+len > i_size)
5347                 len = i_size-off;
5348         toread = len;
5349         while (toread > 0) {
5350                 tocopy = sb->s_blocksize - offset < toread ?
5351                                 sb->s_blocksize - offset : toread;
5352                 bh = ext4_bread(NULL, inode, blk, 0);
5353                 if (IS_ERR(bh))
5354                         return PTR_ERR(bh);
5355                 if (!bh)        /* A hole? */
5356                         memset(data, 0, tocopy);
5357                 else
5358                         memcpy(data, bh->b_data+offset, tocopy);
5359                 brelse(bh);
5360                 offset = 0;
5361                 toread -= tocopy;
5362                 data += tocopy;
5363                 blk++;
5364         }
5365         return len;
5366 }
5367
5368 /* Write to quotafile (we know the transaction is already started and has
5369  * enough credits) */
5370 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5371                                 const char *data, size_t len, loff_t off)
5372 {
5373         struct inode *inode = sb_dqopt(sb)->files[type];
5374         ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5375         int err, offset = off & (sb->s_blocksize - 1);
5376         struct buffer_head *bh;
5377         handle_t *handle = journal_current_handle();
5378
5379         if (EXT4_SB(sb)->s_journal && !handle) {
5380                 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5381                         " cancelled because transaction is not started",
5382                         (unsigned long long)off, (unsigned long long)len);
5383                 return -EIO;
5384         }
5385         /*
5386          * Since we account only one data block in transaction credits,
5387          * then it is impossible to cross a block boundary.
5388          */
5389         if (sb->s_blocksize - offset < len) {
5390                 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5391                         " cancelled because not block aligned",
5392                         (unsigned long long)off, (unsigned long long)len);
5393                 return -EIO;
5394         }
5395
5396         bh = ext4_bread(handle, inode, blk, 1);
5397         if (IS_ERR(bh))
5398                 return PTR_ERR(bh);
5399         if (!bh)
5400                 goto out;
5401         BUFFER_TRACE(bh, "get write access");
5402         err = ext4_journal_get_write_access(handle, bh);
5403         if (err) {
5404                 brelse(bh);
5405                 return err;
5406         }
5407         lock_buffer(bh);
5408         memcpy(bh->b_data+offset, data, len);
5409         flush_dcache_page(bh->b_page);
5410         unlock_buffer(bh);
5411         err = ext4_handle_dirty_metadata(handle, NULL, bh);
5412         brelse(bh);
5413 out:
5414         if (inode->i_size < off + len) {
5415                 i_size_write(inode, off + len);
5416                 EXT4_I(inode)->i_disksize = inode->i_size;
5417                 ext4_mark_inode_dirty(handle, inode);
5418         }
5419         return len;
5420 }
5421
5422 #endif
5423
5424 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5425                        const char *dev_name, void *data)
5426 {
5427         return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5428 }
5429
5430 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5431 static inline void register_as_ext2(void)
5432 {
5433         int err = register_filesystem(&ext2_fs_type);
5434         if (err)
5435                 printk(KERN_WARNING
5436                        "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5437 }
5438
5439 static inline void unregister_as_ext2(void)
5440 {
5441         unregister_filesystem(&ext2_fs_type);
5442 }
5443
5444 static inline int ext2_feature_set_ok(struct super_block *sb)
5445 {
5446         if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5447                 return 0;
5448         if (sb->s_flags & MS_RDONLY)
5449                 return 1;
5450         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5451                 return 0;
5452         return 1;
5453 }
5454 #else
5455 static inline void register_as_ext2(void) { }
5456 static inline void unregister_as_ext2(void) { }
5457 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5458 #endif
5459
5460 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5461 static inline void register_as_ext3(void)
5462 {
5463         int err = register_filesystem(&ext3_fs_type);
5464         if (err)
5465                 printk(KERN_WARNING
5466                        "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5467 }
5468
5469 static inline void unregister_as_ext3(void)
5470 {
5471         unregister_filesystem(&ext3_fs_type);
5472 }
5473
5474 static inline int ext3_feature_set_ok(struct super_block *sb)
5475 {
5476         if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5477                 return 0;
5478         if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5479                 return 0;
5480         if (sb->s_flags & MS_RDONLY)
5481                 return 1;
5482         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5483                 return 0;
5484         return 1;
5485 }
5486 #else
5487 static inline void register_as_ext3(void) { }
5488 static inline void unregister_as_ext3(void) { }
5489 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5490 #endif
5491
5492 static struct file_system_type ext4_fs_type = {
5493         .owner          = THIS_MODULE,
5494         .name           = "ext4",
5495         .mount          = ext4_mount,
5496         .kill_sb        = kill_block_super,
5497         .fs_flags       = FS_REQUIRES_DEV,
5498 };
5499 MODULE_ALIAS_FS("ext4");
5500
5501 static int __init ext4_init_feat_adverts(void)
5502 {
5503         struct ext4_features *ef;
5504         int ret = -ENOMEM;
5505
5506         ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5507         if (!ef)
5508                 goto out;
5509
5510         ef->f_kobj.kset = ext4_kset;
5511         init_completion(&ef->f_kobj_unregister);
5512         ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5513                                    "features");
5514         if (ret) {
5515                 kfree(ef);
5516                 goto out;
5517         }
5518
5519         ext4_feat = ef;
5520         ret = 0;
5521 out:
5522         return ret;
5523 }
5524
5525 static void ext4_exit_feat_adverts(void)
5526 {
5527         kobject_put(&ext4_feat->f_kobj);
5528         wait_for_completion(&ext4_feat->f_kobj_unregister);
5529         kfree(ext4_feat);
5530 }
5531
5532 /* Shared across all ext4 file systems */
5533 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5534 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5535
5536 static int __init ext4_init_fs(void)
5537 {
5538         int i, err;
5539
5540         ext4_li_info = NULL;
5541         mutex_init(&ext4_li_mtx);
5542
5543         /* Build-time check for flags consistency */
5544         ext4_check_flag_values();
5545
5546         for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5547                 mutex_init(&ext4__aio_mutex[i]);
5548                 init_waitqueue_head(&ext4__ioend_wq[i]);
5549         }
5550
5551         err = ext4_init_es();
5552         if (err)
5553                 return err;
5554
5555         err = ext4_init_pageio();
5556         if (err)
5557                 goto out7;
5558
5559         err = ext4_init_system_zone();
5560         if (err)
5561                 goto out6;
5562         ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5563         if (!ext4_kset) {
5564                 err = -ENOMEM;
5565                 goto out5;
5566         }
5567         ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5568
5569         err = ext4_init_feat_adverts();
5570         if (err)
5571                 goto out4;
5572
5573         err = ext4_init_mballoc();
5574         if (err)
5575                 goto out2;
5576         else
5577                 ext4_mballoc_ready = 1;
5578         err = init_inodecache();
5579         if (err)
5580                 goto out1;
5581         register_as_ext3();
5582         register_as_ext2();
5583         err = register_filesystem(&ext4_fs_type);
5584         if (err)
5585                 goto out;
5586
5587         return 0;
5588 out:
5589         unregister_as_ext2();
5590         unregister_as_ext3();
5591         destroy_inodecache();
5592 out1:
5593         ext4_mballoc_ready = 0;
5594         ext4_exit_mballoc();
5595 out2:
5596         ext4_exit_feat_adverts();
5597 out4:
5598         if (ext4_proc_root)
5599                 remove_proc_entry("fs/ext4", NULL);
5600         kset_unregister(ext4_kset);
5601 out5:
5602         ext4_exit_system_zone();
5603 out6:
5604         ext4_exit_pageio();
5605 out7:
5606         ext4_exit_es();
5607
5608         return err;
5609 }
5610
5611 static void __exit ext4_exit_fs(void)
5612 {
5613         ext4_destroy_lazyinit_thread();
5614         unregister_as_ext2();
5615         unregister_as_ext3();
5616         unregister_filesystem(&ext4_fs_type);
5617         destroy_inodecache();
5618         ext4_exit_mballoc();
5619         ext4_exit_feat_adverts();
5620         remove_proc_entry("fs/ext4", NULL);
5621         kset_unregister(ext4_kset);
5622         ext4_exit_system_zone();
5623         ext4_exit_pageio();
5624         ext4_exit_es();
5625 }
5626
5627 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5628 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5629 MODULE_LICENSE("GPL");
5630 module_init(ext4_init_fs)
5631 module_exit(ext4_exit_fs)